Qwen-3-Nemotron-32B-Reward

71

16

32.0B

1 language

—

by

nvidia

Language Model

OTHER

32B params

New

71 downloads

Early-stage

Try on Hugging Face Add to Compare

Edge AI:

Mobile

Laptop

Server

72GB+ RAM

Mobile

Laptop

Server

Quick Summary

AI model with specialized capabilities.

Device Compatibility

Mobile

4-6GB RAM

Laptop

16GB RAM

Server

GPU

Minimum Recommended

30GB+ RAM

Code Examples

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Example quality - note that higher scores means higher quality, and scores can be negative.pythontransformers

import torch
from transformers import AutoModelForSequenceClassification, AutoTokenizer

model_name = "nvidia/Qwen-3-Nemotron-32B-Reward"

model = AutoModelForSequenceClassification.from_pretrained(model_name, torch_dtype=torch.float16, device_map="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)

prompt = "What is 1+1?"
good_response = "1+1=2"
bad_response = "1+1=3"

for response in [good_response, bad_response]:
    messages = [{'role': "user", "content": prompt + " /nothink"}, {'role': "assistant", "content": "<think>\n\n</think>\n\n" + response}]
    tokenized_message = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=False, return_tensors="pt", return_dict=True)
    reward = model(tokenized_message['input_ids'].cuda(),attention_mask=tokenized_message['attention_mask'].cuda()).logits[0][0].item()
    print(reward)

# Example quality - note that higher scores means higher quality, and scores can be negative.

# reward for good_response = 8.0234375
# reward for bad_response = -7.9765625

Deploy This Model

Production-ready deployment in minutes

Together.ai

Instant API access to this model

Fastest API

Production-ready inference API. Start free, scale to millions.

Replicate

One-click model deployment

Easiest Setup

Run models in the cloud with simple API. No DevOps required.

Disclosure: We may earn a commission from these partners. This helps keep LLMYourWay free.