gpt-6POST
OpenAI's internal model solves nearly 50% of math benchmark, far ahead of Astra
A leak says the internal OpenAI model behind the Navier-Stokes breakthrough solved nearly 50% of OpenAI's advanced internal math benchmark, versus 10% for GPT-6 Astra.
MTS
@MTSliveArchived 2026.09.14Originally posted 2026.09.09
MTS @MTSliveOriginally posted 2026.09.09
SITUATION DETECTED: The internal OpenAI model behind the Navier-Stokes breakthrough has shown a massive leap in capability. On OpenAI’s advanced internal mathematics benchmark, the new model solved nearly 50% of the problems, compared to just 10% from GPT-6 Astra.
x.com/mtslive/status/2097391968708108295Archived 2026.09.14 15:41