News on Medial

DeepSeek claims its 'reasoning' model beats OpenAI's o1 on certain benchmarks | TechCrunch

TechCrunch · 2m

DeepSeek, a Chinese AI lab, has unveiled DeepSeek-R1, claiming it performs comparably to OpenAI’s o1 on benchmarks like AIME and SWE-bench Verified. Released on Hugging Face under MIT license, R1 is commercially usable with 671 billion parameters, offering more reliable reasoning in fields like physics and math. Distilled versions are also available. However, R1 faces content restrictions by China's regulator. This release coincides with potential tighter U.S. export rules for AI technology to China.