传 OpenAI 内部模型 Bel 解出纳维-斯托克斯难题
爆料称 OpenAI 内部模型「Bel」比 GPT-6 Astra 更强，据称解出悬置约 90 年的纳维-斯托克斯千禧年难题，但结果有争议、署名也起纠纷。
🚨 Bel Leaks: OpenAI's Internal Model Beyond Astra >OpenAI officially confirmed an internal model "significantly more capable than GPT-6 Astra" solved >the Navier-Stokes Millennium Prize Problem >One of the seven deepest unsolved problems in mathematics, open for ~90 years >OpenAI isn't claiming the $1M prize, framing it instead as a signal of how fast their models are improving >The result is contested an Anthropic researcher was reportedly close to a related solution independently, and there's a credit dispute over authorship >Rumored internal codename for this model: Bel >Could become the base for an AGI-threshold model >Reportedly 10T+ total parameters Timing suggests this could land before 2027, after GPT-6.1 Astra ships