Saxo

This model is based on the Qwen/Qwen2.5-32B-Instruct architecture and is licensed under Apache 2.0.

Open Japanese LLM Leaderboard 🏆 Rank-1 2025/01/12~ AI専門企業であるLinkbricks Horizon-AI のデータサイエンティストであるジ・ユンソン(Saxo)代表が Saxo/Linkbricks-Horizon-AI-Japanese-Superb-V3-70Bベースモデルを使用し、H100-80G 8個を通じて約 日本語 SFT->DPO した日本語強化言語モデル。 3千万件の日本ニュース及びウィキコーパスを基準に、様々なタスク別の日本語・韓国語・中国語・英語クロス学習データと数学及び論理判断データを通じて、日中韓英言語クロス補強処理と複雑な論理問題にも対応できるように訓練したモデルです。 -トークナイザーは、単語拡張なしでベースモデルのまま使用します。 -カスタマーレビューやソーシャル投稿の高次元分析及びコーディングとライティング、数学、論理判断などが強化されたモデル。 -128k-Context Window -Function Call -128k-Context Window -Deepspeed Stage=3、rslora及びBAdam Layer Modeを使用 -「transformersversion」: 「4.46.3」 AI 전문 기업인 Linkbricks Horizon-AI 의 데이터사이언티스트인 지윤성(Saxo) 대표가 Saxo/Linkbricks-Horizon-AI-Japanese-Superb-V3-70B 베이스모델을 사용해서 H100-80G 8개를 통해 일본어 SFT->DPO 한 일본어 강화 언어 모델 3천만건의 일본 뉴스 및 위키 코퍼스를 기준으로 다양한 테스크별 일본어-한국어-중국어-영어 교차 학습 데이터와 수학 및 논리판단 데이터를 통하여 한중일영 언어 교차 증강 처리와 복잡한 논리 문제 역시 대응 가능하도록 훈련한 모델이다. -토크나이저는 단어 확장 없이 베이스 모델 그대로 사용 -고객 리뷰나 소셜 포스팅 고차원 분석 및 코딩과 작문, 수학, 논리판단 등이 강화된 모델 -128k-Context Window -Function Call 및 Tool Calling 지원 -128k-Context Window -Deepspeed Stage=3, rslora 및 BAdam Layer Mode 사용 -"transformersversion": "4.46.3" Finetuned by Mr. Yunsung Ji (Saxo), a data scientist and CEO at Linkbricks Horiozn-AI, a company specializing in AI and big data analytics Japanese SFT->DPO training model based on Saxo/Linkbricks-Horizon-AI-Japanese-Superb-V3-70B through 8 H100-80Gs as a Japanese boosting language model It is a model that has been trained to handle Japanese-Korean-Chinese-English cross-training data and 30M Japanese news corpus and logic judgment data for various tasks to enable cross-fertilization processing and complex Korean logic & math problems. -Tokenizer uses the base model without word expansion -Models enhanced with high-dimensional analysis of customer reviews and social posts, as well as coding, writing, math and decision making -Function Calling -128k-Context Window -Deepspeed Stage=3, use rslora and BAdam Layer Mode www.horizonai.ai , www.linkbricks.com , www.linkbricks.vc

This model is based on the Google Gemma 2 27B IT architecture and is licensed under Apache 2.0.

This model is based on the Mistral Small Instruct 2409 architecture and is licensed under Apache 2.0.

This model is based on the Saxo/Linkbricks-Horizon-AI-Korean-Avengers-V1-27B and is licensed under Apache 2.0.

This model is based on the Linkbricks Horizon AI Avengers V5 32B and is licensed under Apache 2.0. It utilizes the transformers library.

This model is based on the Linkbricks Horizon AI Korean Avengers V2 27B and is licensed under the Apache 2.0 license, utilizing the transformers library.

This model is based on the Linkbricks Horizon AI Avengers V1 32B and is licensed under Apache 2.0. It utilizes the transformers library.

This model is based on the Linkbricks Horizon AI Avengers V3 32B and is licensed under Apache 2.0. It utilizes the transformers library.

This model is based on the Linkbricks Horizon AI Avengers V2 32B and is licensed under Apache 2.0. It utilizes the transformers library.

A library for natural language processing using the transformers framework. It is licensed under Apache 2.0 and is based on the Google Gemma 2 27B IT model.

AIとビッグデータ分析の専門企業であるLinkbricksのデータサイエンティストであるジ・ユンソン(Saxo)ディレクターが Hermes-3-Llama-3.1-70Bベースモデルを使用し、H100-80G 8個を通じて約20%程度のパラメータを日本語CPT(Continued-Pretraining)->SFT->DPOした日本語強化言語モデル。 2千万件の日本ニュース及びウィキコーパスを基準に、様々なタスク別の日本語・韓国語・中国語・英語クロス学習データと数学及び論理判断データを通じて、日中韓英言語クロス補強処理と複雑な論理問題にも対応できるように訓練したモデルです。 -トークナイザーは、単語拡張なしでベースモデルのまま使用します。 -カスタマーレビューやソーシャル投稿の高次元分析及びコーディングとライティング、数学、論理判断などが強化されたモデル。 -128k-Context Window -Function Call -128k-Context Window -Deepspeed Stage=3、rslora及びBAdam Layer Modeを使用 -「transformersversion」: 「4.46.3」 AI 와 빅데이터 분석 전문 기업인 Linkbricks의 데이터사이언티스트인 지윤성(Saxo) 이사가 Hermes-3-Llama-3.1-70B 베이스모델을 사용해서 H100-80G 8개를 통해 약 20%정도의 파라미터를 일본어 CPT(Continued-Pretraining)->SFT->DPO 한 일본어 강화 언어 모델 2천만건의 일본 뉴스 및 위키 코퍼스를 기준으로 다양한 테스크별 일본어-한국어-중국어-영어 교차 학습 데이터와 수학 및 논리판단 데이터를 통하여 한중일영 언어 교차 증강 처리와 복잡한 논리 문제 역시 대응 가능하도록 훈련한 모델이다. -토크나이저는 단어 확장 없이 베이스 모델 그대로 사용 -고객 리뷰나 소셜 포스팅 고차원 분석 및 코딩과 작문, 수학, 논리판단 등이 강화된 모델 -128k-Context Window -Function Call 및 Tool Calling 지원 -128k-Context Window -Deepspeed Stage=3, rslora 및 BAdam Layer Mode 사용 -"transformersversion": "4.46.3" Finetuned by Mr. Yunsung Ji (Saxo), a data scientist at Linkbricks, a company specializing in AI and big data analytics about 20% of total parameters Japanese CPT(Continued-Pretraining)->SFT->DPO training model based on Hermes-3-Llama-3.1-70B through 8 H100-80Gs as a Japanese boosting language model It is a model that has been trained to handle Japanese-Korean-Chinese-English cross-training data and 20M Japanese news corpus and logic judgment data for various tasks to enable cross-fertilization processing and complex Korean logic & math problems. -Tokenizer uses the base model without word expansion -Models enhanced with high-dimensional analysis of customer reviews and social posts, as well as coding, writing, math and decision making -Function Calling -128k-Context Window -Deepspeed Stage=3, use rslora and BAdam Layer Mode

This model is based on the Linkbricks Horizon AI Avengers V4 32B and is licensed under the Apache 2.0 license, utilizing the transformers library.