[Brief] [Brief] DeepSeek Narrows Gap with Top U.S. AI Models to 3%, Confirmed by Multiple Analysts

Multiple AI benchmark and analysis firms published reports on October 4 indicating the capability gap between new DeepSeek models and top U.S. AI models has narrowed to within 3% across multiple core benchmarks. Analysts noted the narrowing was primarily driven by Chinese models’ rapid improvements in reasoning, code generation and math tasks.

The U.S. AI Safety Institute stated in a brief that the rate at which the Sino-U.S. AI gap has narrowed has exceeded prior expectations, recommending federal AI policy makers accelerate the actual enforcement cadence of AI chip export controls.

Editor: GoodInfo Global News Team