ZeroWw

191 models • 1 total models in database
Sort by:

NeuralDaredevil-8B-abliterated-GGUF

NaNK
license:mit
209
4

Test

license:mit
204
0

Meta-Llama-3.1-8B-Instruct-abliterated-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
183
17

llama3-8B-DarkIdol-2.2-Uncensored-1048K-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
134
7

Llama-3-8B-Lexi-Uncensored-GGUF

NaNK
license:mit
124
3

NSFW_DPO_Noromaid-7b-Mistral-7B-Instruct-v0.1-GGUF

NaNK
license:mit
113
3

GLM-Z1-9B-0414-GGUF

NaNK
license:mit
106
0

Mistral-Nemo-Base-2407-GGUF

license:mit
104
5

DarkIdol-Llama-3.1-8B-Instruct-1.0-Uncensored-GGUF

NaNK
license:mit
102
1

Llama-3.2-3B-Instruct-abliterated-GGUF

NaNK
license:mit
97
2

TwinLlama-3.1-8B-GGUF

NaNK
license:mit
96
0

Phi-4-mini-instruct-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

license:mit
96
0

Llama-3.1-Storm-8B-GGUF

NaNK
license:mit
94
0

NuminaMath-7B-TIR-GGUF

NaNK
license:mit
91
3

Gemmasutra-Mini-2B-v1-GGUF

NaNK
license:mit
84
2

llama3-8B-DarkIdol-2.1-Uncensored-32K-GGUF

NaNK
license:mit
81
1

Phi-3-mini-4k-instruct-GGUF

license:mit
80
0

Qwen3-8B-abliterated-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
80
0

Llama-3.2-1B-Instruct-GGUF

NaNK
license:mit
78
0

Qwen3-8B-GGUF

NaNK
license:mit
76
2

Pythia-Chat-Base-7B-GGUF

NaNK
license:mit
75
1

Meta-Llama-3.1-8B-Claude-39fail-3000total-GGUF

NaNK
license:mit
74
1

Phi-3.5-mini-instruct_Uncensored-GGUF

license:mit
74
1

L3-Aethora-15B-V2-GGUF

NaNK
license:mit
73
2

Llama3.1-8B-Enigma-GGUF

NaNK
license:mit
70
3

gemma-3-4b-it-abliterated-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
69
4

Celeste-12B-V1.6-GGUF

NaNK
license:mit
67
0

Qwen3-4B-abliterated-GGUF

NaNK
license:mit
67
0

CodeQwen1.5-7B-Chat-GGUF

NaNK
license:mit
66
0

Replete-LLM-Qwen2-7b_Beta-Preview-GGUF

NaNK
license:mit
62
0

ghost-8b-beta-1608-GGUF

NaNK
license:mit
62
0

Hunyuan-7B-Instruct-GGUF

NaNK
license:mit
62
0

Mistral-NeMo-Minitron-8B-Instruct-GGUF

NaNK
license:mit
61
0

gemma-2-2b-it-GGUF

NaNK
license:mit
60
0

phi3-uncensored-chat-GGUF

license:mit
60
0

Qwen3-4B-Thinking-2507-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
59
0

Qwen3-4B-Esper3-GGUF

NaNK
license:mit
58
0

aya-23-8B-GGUF

NaNK
license:mit
57
1

gpt2-xl-GGUF

license:mit
57
0

Mistral-Nemo-Instruct-2407-GGUF

license:mit
56
1

Meta-Llama-3.1-8B-Instruct-GGUF

NaNK
license:mit
56
0

Qwen3-4B-GGUF

NaNK
license:mit
54
2

Llama3.1-8B-ShiningValiant2-GGUF

NaNK
license:mit
54
1

ghost-8b-beta-GGUF

NaNK
license:mit
51
1

Art-0-8B-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
51
0

glm-4-9b-chat-GGUF

NaNK
license:mit
50
4

Llama-3.1-Minitron-4B-Width-Base-GGUF

NaNK
license:mit
50
3

Phi-4-mini-reasoning-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

license:mit
49
0

glm-4-9b-chat-1m-GGUF

NaNK
license:mit
48
1

Qwen2.5-1.5B-Instruct-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
48
0

Meta-Llama-3-8B-Instruct-GGUF

NaNK
license:mit
47
0

L3.2-Rogue-Creative-Instruct-Uncensored-Abliterated-7B-D_AU-SILLY

NaNK
license:mit
45
1

Phi-3-medium-128k-instruct-GGUF

license:mit
43
3

Hermes-2-Pro-Llama-3-8B-GGUF

NaNK
license:mit
42
0

L3-8b-Rosier-v1-GGUF

NaNK
license:mit
42
0

Mistral-7B-Instruct-v0.3-GGUF

NaNK
license:mit
42
0

llama-3-Nephilim-v3-8B-GGUF

NaNK
license:mit
42
0

Lumimaid-v0.2-12B-GGUF

NaNK
license:mit
41
2

Symbol-LLM-8B-Instruct-GGUF

NaNK
license:mit
41
0

codegeex4-all-9b-GGUF

NaNK
license:mit
40
5

microsoft_WizardLM-2-7B-GGUF

NaNK
license:mit
40
2

Seed-Coder-8B-Reasoning-GGUF

NaNK
license:mit
40
0

Smegmma-9B-v1-GGUF

NaNK
license:mit
39
3

Llama-3-8B-Instruct-Gradient-4194k-GGUF

NaNK
license:mit
39
1

llama3-turbcat-instruct-8b-GGUF

NaNK
license:mit
39
1

gemma-2-2b-it-abliterated-GGUF

NaNK
license:mit
39
0

Qwen3-4B-Instruct-2507-GGUF

NaNK
license:mit
39
0

Mistral-7B-Instruct-v0.3-SILLY

NaNK
license:mit
38
3

L3-8B-Celeste-v1-GGUF

NaNK
license:mit
38
1

Lumimaid-v0.2-8B-GGUF

NaNK
license:mit
38
1

Arcee-Spark-GGUF

license:mit
38
0

L3-SthenoMaid-8B-V1-GGUF

NaNK
license:mit
38
0

Llama-3-8B-Instruct-Gradient-1048k-GGUF

NaNK
license:mit
37
2

Phi-3-mini-4k-geminified-GGUF

license:mit
37
0

L3-8B-Celeste-V1.2-GGUF

NaNK
license:mit
37
0

L3.1-8B-Celeste-V1.5-GGUF

NaNK
license:mit
37
0

Llama-3.2-3B-Instruct-GGUF

NaNK
license:mit
37
0

aya-expanse-8b-GGUF

NaNK
license:mit
36
1

Qwen3-8B-Esper3-GGUF

NaNK
license:mit
36
0

Hathor_Stable-v0.2-L3-8B-GGUF

NaNK
license:mit
35
0

Meta-Llama-3-8B-Instruct-abliterated-v3-GGUF

NaNK
license:mit
35
0

L3-8B-Stheno-v3.3-32K-GGUF

NaNK
license:mit
34
2

Gemma-2-9B-It-SPPO-Iter3-GGUF

NaNK
license:mit
34
2

Gemmasutra-9B-v1b-GGUF

NaNK
license:mit
34
1

shieldgemma-2b-GGUF

NaNK
license:mit
34
0

L3-Blackfall-Summanus-v0.1-15B-GGUF

NaNK
license:mit
33
1

cogito-v1-preview-llama-8B-GGUF

NaNK
license:mit
33
0

DeepSeek-Coder-V2-Lite-Base-GGUF

license:mit
32
0

LLaMAX3-8B-Alpaca-GGUF

NaNK
license:mit
32
0

Phi-3-mini-128k-instruct-GGUF

license:mit
31
2

DeepSeek-V2-Lite-Chat-GGUF

license:mit
31
1

LLaMAX3-8B-GGUF

NaNK
license:mit
31
0

Tiger-Gemma-9B-v1-GGUF

NaNK
license:mit
31
0

neural-chat-7b-v3-3-GGUF

NaNK
license:mit
28
0

Gemma-3-R1-4B-v1-GGUF

NaNK
license:mit
27
0

gemma-2-9b-it-GGUF

NaNK
license:mit
26
1

internlm2_5-7b-chat-GGUF

NaNK
license:mit
25
2

Qwen1.5-7B-Chat-GGUF

NaNK
license:mit
25
0

Smegmma-Deluxe-9B-v1-GGUF

NaNK
license:mit
24
0

TwinLlama-3.1-8B-SILLY

NaNK
license:mit
24
0

Qwen2.5-3B-Instruct-GGUF

NaNK
license:mit
24
0

Samantha-Qwen-2-7B-GGUF

NaNK
license:mit
23
0

Phi-3-song-lyrics-1.0-GGUF

license:mit
23
0

DarkIdol-Llama-3.1-8B-Instruct-1.0-Uncensored-SILLY

NaNK
license:mit
22
1

h2ogpt-4096-llama2-13b-chat-GGUF

NaNK
license:mit
22
0

ghost-8b-beta-1608-SILLY

NaNK
license:mit
21
0

gemma-3-12b-it-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
21
0

granite-3.3-8b-instruct-GGUF

NaNK
license:mit
21
0

open_llama_7b_v2-GGUF

NaNK
license:mit
20
0

ghost-7b-alpha-GGUF

NaNK
license:mit
20
0

xLAM-1b-fc-r-GGUF

NaNK
license:mit
20
0

xLAM-7b-fc-r-GGUF

NaNK
license:mit
20
0

Qwen2.5-7B-Instruct-GGUF

NaNK
license:mit
20
0

EuroLLM-1.7B-Instruct-GGUF

NaNK
license:mit
20
0

internlm2_5-7b-chat-1m-GGUF

NaNK
license:mit
19
1

Mixtral_AI_Cyber_4.0-GGUF

license:mit
19
1

Gemma-2-9B-It-SPPO-Iter3-SILLY

NaNK
license:mit
19
0

open_llama_3b_v2-GGUF

NaNK
license:mit
17
0

gemma-2-2b-it-SILLY

NaNK
license:mit
17
0

Gemmasutra-Mini-2B-v1-SILLY

NaNK
license:mit
17
0

Mistral-Nemo-12B-ArliAI-RPMax-v1.2-GGUF

NaNK
license:mit
17
0

Llama3.1-8B-Enigma-SILLY

NaNK
license:mit
16
1

Moistral-11B-v3-GGUF

NaNK
license:mit
16
0

ghost-8b-beta-SILLY

NaNK
license:mit
16
0

Meta-Llama-3.1-8B-Instruct-abliterated-SILLY

NaNK
license:mit
16
0

Gemmasutra-9B-v1b-SILLY

NaNK
license:mit
16
0

Lumimaid-v0.2-12B-SILLY

NaNK
license:mit
16
0

Llama-3.1-Storm-8B-SILLY

NaNK
license:mit
16
0

Mistral-Nemo-Instruct-2407-SILLY

license:mit
15
0

L3.1-8B-Celeste-V1.5-SILLY

NaNK
license:mit
15
0

Llama3.1-8B-ShiningValiant2-SILLY

NaNK
license:mit
15
0

Phi-3.5-mini-instruct-GGUF

license:mit
15
0

amoral-gemma3-4B-GGUF

NaNK
license:mit
15
0

DeepSeek-R1-0528-Qwen3-8B-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
15
0

gemma-3-270m-it-GGUF

license:mit
15
0

Marco-o1-GGUF

license:mit
14
0

DeepSeek-R1-Distill-Qwen-7B-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
14
0

Josiefied-Qwen3-8B-abliterated-v1-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
14
0

Meta-Llama-3.1-8B-Instruct-SILLY

NaNK
license:mit
13
1

Llama-3.2-3B-Instruct-abliterated-SILLY

NaNK
license:mit
13
1

Seed-Coder-8B-Instruct-GGUF

NaNK
license:mit
13
1

Mistroll-7B-v2.2-GGUF

NaNK
license:mit
13
0

phillama-3.8b-v0.1-GGUF

NaNK
license:mit
13
0

h2o-danube3-500m-chat-GGUF

license:mit
13
0

h2o-danube3-4b-chat-GGUF

NaNK
license:mit
13
0

Lumimaid-v0.2-8B-SILLY

NaNK
license:mit
13
0

OpenELM-3B-Instruct-GGUF

NaNK
license:mit
13
0

palmer-004-turbo-GGUF

license:mit
12
1

Phi-3-mini-128k-instruct-abliterated-v3-GGUF

license:mit
12
0

Replete-LLM-Qwen2-7b_Beta-Preview-SILLY

NaNK
license:mit
12
0

aya-expanse-8b-SILLY

NaNK
license:mit
12
0

gemma-2-2b-it-abliterated-SILLY

NaNK
license:mit
11
2

NeuralPipe-7B-slerp-GGUF

NaNK
license:mit
11
0

SOLAR-10.7B-Instruct-v1.0-GGUF

NaNK
license:mit
11
0

Phi-3.5-mini-instruct-SILLY

license:mit
11
0

Llama-3.2-1B-Instruct-SILLY

NaNK
license:mit
11
0

gemma-3-1b-it-abliterated-GGUF

NaNK
license:mit
11
0

Yi-1.5-9B-32K-GGUF

NaNK
license:mit
10
1

Yi-1.5-6B-Chat-GGUF

NaNK
license:mit
10
1

Llama-3.2-3B-Instruct-SILLY

NaNK
license:mit
10
0

Qwen2.5-1.5B-Instruct-SILLY

ZeroWw 'SILLY' version. The original model has been quantized (fq8 version) and a percentage of it's tensors have been modified adding some noise. Full colab: https://colab.research.google.com/drive/1a7seagBzu5l3k3FL4SFk0YJocl7nsDJw?usp=sharing Fast colab: https://colab.research.google.com/drive/1SDD7ox21di82Y9v68AUoy0PhkxwBVvN?usp=sharing Original reddit post: https://www.reddit.com/r/LocalLLaMA/comments/1ec0s8p/imadeasillytest/ I created a program to randomize the weights of a model. The program has 2 parameters: the percentage of weights to modify and the percentage of the original value to randmly apply to each weight. At the end I check the resulting GGUF file for binary differences. In this example I set to modify 100% of the weights of Mistral 7b Instruct v0.3 by a maximum of 15% deviation. Since the deviation is calculated on the F32 weights, when quantized to Q8\0 this changes. So, in the end I got a file that compared to the original has: The cool thing is that chatting with the model I see no apparent difference and the model still works nicely as the original. Since I am running everything on CPU, I could not run perplexity scores or anything computing intensive. As a small test, I asked the model a few questions (like the history of the roman empire) and then fact check its answer using a big model. No errors were detected.

NaNK
license:mit
10
0

internlm3-8b-instruct-GGUF

NaNK
license:mit
10
0

Mistral-Nemo-12B-ArliAI-RPMax-v1.2-SILLY

NaNK
license:mit
9
2

MixTAO-7Bx2-MoE-v8.1-GGUF

NaNK
license:mit
9
1

Qwen2.5-3B-Instruct-SILLY

NaNK
license:mit
9
0

EuroLLM-1.7B-Instruct-SILLY

ZeroWw 'SILLY' version. The original model has been quantized (fq8 version) and a percentage of it's tensors have been modified adding some noise. Full colab: https://colab.research.google.com/drive/1a7seagBzu5l3k3FL4SFk0YJocl7nsDJw?usp=sharing Fast colab: https://colab.research.google.com/drive/1SDD7ox21di82Y9v68AUoy0PhkxwBVvN?usp=sharing Original reddit post: https://www.reddit.com/r/LocalLLaMA/comments/1ec0s8p/imadeasillytest/ I created a program to randomize the weights of a model. The program has 2 parameters: the percentage of weights to modify and the percentage of the original value to randmly apply to each weight. At the end I check the resulting GGUF file for binary differences. In this example I set to modify 100% of the weights of Mistral 7b Instruct v0.3 by a maximum of 15% deviation. Since the deviation is calculated on the F32 weights, when quantized to Q8\0 this changes. So, in the end I got a file that compared to the original has: The cool thing is that chatting with the model I see no apparent difference and the model still works nicely as the original. Since I am running everything on CPU, I could not run perplexity scores or anything computing intensive. As a small test, I asked the model a few questions (like the history of the roman empire) and then fact check its answer using a big model. No errors were detected.

NaNK
license:mit
9
0

Moistral-11B-v4-GGUF

NaNK
license:mit
8
0

Phi-3.5-mini-instruct_Uncensored-SILLY

license:mit
8
0

Phi3Unlocked-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

license:mit
8
0

Mistral-NeMo-Minitron-8B-Instruct-SILLY

NaNK
license:mit
8
0

Qwen3-0.6B-GGUF

NaNK
license:mit
7
1

Falcon-H1-7B-Instruct-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
7
0

Qwen2.5-7B-Instruct-SILLY

NaNK
license:mit
6
1

granite-3.1-8b-instruct-GGUF

NaNK
license:mit
6
0

granite-3.1-8b-instruct-abliterated-GGUF

NaNK
license:mit
6
0

SOLAR-10.7B-Instruct-v1.0-SILLY

NaNK
license:mit
4
0

neural-chat-7b-v3-3-SILLY

NaNK
license:mit
4
0

gemma-3-1b-it-GGUF

NaNK
license:mit
4
0

Marco-o1-SILLY

ZeroWw 'SILLY' version. The original model has been quantized (fq8 version) and a percentage of it's tensors have been modified adding some noise. Full colab: https://colab.research.google.com/drive/1a7seagBzu5l3k3FL4SFk0YJocl7nsDJw?usp=sharing Fast colab: https://colab.research.google.com/drive/1SDD7ox21di82Y9v68AUoy0PhkxwBVvN?usp=sharing Original reddit post: https://www.reddit.com/r/LocalLLaMA/comments/1ec0s8p/imadeasillytest/ I created a program to randomize the weights of a model. The program has 2 parameters: the percentage of weights to modify and the percentage of the original value to randmly apply to each weight. At the end I check the resulting GGUF file for binary differences. In this example I set to modify 100% of the weights of Mistral 7b Instruct v0.3 by a maximum of 15% deviation. Since the deviation is calculated on the F32 weights, when quantized to Q8\0 this changes. So, in the end I got a file that compared to the original has: The cool thing is that chatting with the model I see no apparent difference and the model still works nicely as the original. Since I am running everything on CPU, I could not run perplexity scores or anything computing intensive. As a small test, I asked the model a few questions (like the history of the roman empire) and then fact check its answer using a big model. No errors were detected.

license:mit
3
0

granite-3.1-3b-a800m-instruct-GGUF

NaNK
license:mit
3
0

Llama-Deepsync-1B-GGUF

NaNK
license:mit
3
0

EXAONE-Deep-2.4B-GGUF

NaNK
license:mit
3
0

Phi3Unlocked-SILLY

license:mit
2
0

granite-3.1-2b-instruct-GGUF

NaNK
license:mit
2
0

EXAONE-Deep-7.8B-GGUF

NaNK
license:mit
2
0

Llama-3.1-Nemotron-Nano-8B-v1-GGUF

NaNK
license:mit
2
0

GLM-4-9B-0414-GGUF

NaNK
license:mit
1
1

granite-3.2-2b-instruct-GGUF

NaNK
license:mit
1
0

medgemma-4b-it-GGUF

My own (ZeroWw) quantizations. output and embed tensors quantized to f16. all other tensors quantized to q5k or q6k. Result: both f16.q6 and f16.q5 are smaller than q80 standard quantization and they perform as well as the pure f16.

NaNK
license:mit
1
0

OpenThinker3-7B-GGUF

NaNK
license:mit
1
0

L3-Nymeria-15B-GGUF

NaNK
0
1