DeepSeek has launched DeepSeek-R1, an open-source reasoning model available on Hugging Face. It outperforms OpenAI's o1 in benchmarks like AIME, MATH-500, and SWE-bench Verified. R1 incorporates 671 billion parameters, enabling superior problem-solving, especially in complex subjects like physics and math. Distilled versions ranging from 1.5 billion to 70 billion parameters are also available for broader accessibility, including use on personal laptops. The model is 90%-95% cheaper than OpenAI's offerings, prompting developers to create over 500 derivative models with 2.5 million downloads in record time.
DeepSeek-R1 is a new reasoning model that competes with OpenAI's o1, available openly for commercial use and optimized for complex reasoning tasks.
Collection
[
|
...
]