GPT-6.1 Sol 以五分之一成本对标 Astra

Choblin @choblin29原推发布于 2026.09.30

在 X 上打开原推

🚨HOLY SHIT. GPT-6.1 Sol is putting up Astra and Opus 5.5-agent numbers at a FRACTION of the cost. This model is a BEAST: >matches GPT-6 Astra on DeepSWE >at roughly ONE-FIFTH the cost >beats GPT-6 Sol's best DeepSWE score by 6.4 points >beats Opus 5.5 on GDP.pdf >at less than HALF the cost >beats Opus 5.5 by 2.2 points on AutomationBench >at roughly ONE-THIRD the cost >comes within 2.1 points of Astra on OSWorld >at roughly ONE-SEVENTH the cost >scientific tasks cost just $5.47 >Opus 5.5: $23.21 >Astra: $23.80 And on Terminal-Bench Science, 6.1 Sol more than DOUBLES GPT-6 Sol's score. The frontier price war is getting ridiculous.