@deepseek_ai - DeepSeek
๐ DeepSeek-R1 is here! โก Performance on par with OpenAI-o1 ๐ Fully open-source model & technical report ๐ MIT licensed: Distill & commercialize freely! ๐ Website & API are live now! Try DeepThink at http://chat.deepseek.com today! ๐ 1/n
@deepseek_ai - DeepSeek
๐ฅ Bonus: Open-Source Distilled Models! ๐ฌ Distilled from DeepSeek-R1, 6 small models fully open-sourced ๐ 32B & 70B models on par with OpenAI-o1-mini ๐ค Empowering the open-source community ๐ Pushing the boundaries of **open AI**! ๐ 2/n
@deepseek_ai - DeepSeek
๐ License Update! ๐ DeepSeek-R1 is now MIT licensed for clear open access ๐ Open for the community to leverage model weights & outputs ๐ ๏ธ API outputs can now be used for fine-tuning & distillation ๐ 3/n
@deepseek_ai - DeepSeek
๐ ๏ธ DeepSeek-R1: Technical Highlights ๐ Large-scale RL in post-training ๐ Significant performance boost with minimal labeled data ๐ข Math, code, and reasoning tasks on par with OpenAI-o1 ๐ More details: https://github.com/deepseek-ai/DeepSeek-R1/blob/main/DeepSeek_R1.pdf ๐ 4/n
@deepseek_ai - DeepSeek
๐ API Access & Pricing โ๏ธ Use DeepSeek-R1 by setting model=deepseek-reasoner ๐ฐ $0.14 / million input tokens (cache hit) ๐ฐ $0.55 / million input tokens (cache miss) ๐ฐ $2.19 / million output tokens ๐ API guide: https://api-docs.deepseek.com/guides/reasoning_model ๐ 5/n