<br>[DeepSeek open-sourced](https://gitlab.kitware.com) DeepSeek-R1, an LLM fine-tuned with reinforcement learning (RL) to [improve reasoning](http://62.234.223.2383000) ability. DeepSeek-R1 attains outcomes on par with OpenAI's o1 model on a number of criteria, including MATH-500 and [SWE-bench](http://118.31.167.22813000).<br>
<br>DeepSeek-R1 is based upon DeepSeek-V3, a mixture of experts (MoE) model recently [open-sourced](http://60.250.156.2303000) by DeepSeek. This base model is fine-tuned using Group [Relative Policy](https://tricityfriends.com) Optimization (GRPO), a [reasoning-oriented](https://career.agricodeexpo.org) version of RL. The research group also performed knowledge distillation from DeepSeek-R1 to open-source Qwen and Llama models and [launched numerous](https://collegetalks.site) versions of each
