Astra 打满 FrontierMath Tier 4

Haider. @haider1原推发布于 2026.09.11

在 X 上打开原推

GPT-6 Astra has saturated the FrontierMath Tier 4 benchmark these are research-level math problems designed to take expert mathematicians hours or even days to solve -- and somehow the benchmark reached saturation in roughly 14 months way beyond what i expected

这条是引用着另一条写的：x.com/epochairesearch/status/2098103831502708864