AI BriefingDeepSeek-R1 Released with Technical Report
Key point
DeepSeek has released DeepSeek-R1, an open-source reasoning model with performance on par with OpenAI-o1.
Details
DeepSeek-R1 is a new reasoning model that shows performance on par with OpenAI-o1 on math, code, and reasoning tasks. The model is released under the MIT License, allowing free use of the model weights and outputs, including commercial use.
DeepSeek has also released 6 smaller distilled models based on DeepSeek-R1. In particular, the 32B and 70B models deliver performance similar to OpenAI-o1-mini, supporting the advancement of the open-source community.
Technically, large-scale reinforcement learning (RL) was applied at the post-training stage, maximizing performance with only minimal labeled data. The API can be used via the model=deepseek-reasoner setting, and is offered at a low cost of about $0.14–$0.55 per 1 million input tokens.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.