• 本文
• リロード
• 差分
• バックアップ
• 名前変更
• リファラ
• 新規ページ
• 編集

DeepSeek

DeepSeek-R1

Top / DeepSeek-R1

https://github.com/deepseek-ai/DeepSeek-R1

• DeepSeek-R1-Zero?	671B	669.28GB (BF16)
  DeepSeek-R1	671B	656.38GB (BF16)

DeepSeek-R1-0528

• DeepSeek-R1-0528-Qwen3-8B

DeepSeek-R1-Distill

• DeepSeek-R1-Distill models are fine-tuned based on open-source models, using samples generated by DeepSeek-R1. We slightly change their configs and tokenizers. Please use our setting to run these models.

  Model	BF16	Base Model
  DeepSeek-R1-Distill-Qwen-1.5B	3.55GB	Qwen2.5-Math-1.5B
  DeepSeek-R1-Distill-Qwen-7B	15.23GB	Qwen2.5-Math-7B?
  DeepSeek-R1-Distill-Llama-8B	16.06GB	Llama-3.1-8B?
  DeepSeek-R1-Distill-Qwen-14B	29.54GB	Qwen2.5-14B
  DeepSeek-R1-Distill-Qwen-32B	65.56GB	Qwen2.5-32B
  DeepSeek-R1-Distill-Llama-70B	136.87GB	Llama-3.3-70B-Instruct

メモ

• DeepSeek-R1-Distill-Qwen-32BはさまざまなベンチマークでOpenAI o1-mini?を上回った

テクニカルレポート

• https://github.com/deepseek-ai/DeepSeek-R1/blob/23807ce/DeepSeek_R1.pdf

関連

• DeepSeek-R1-Zero?

• CoT

参考

• https://zenn.dev/asap/articles/34237ad87f8511
• https://yoheikikuta.github.io/blog/2025-01-23-deepseekr1_techinical_paper/