观点：OpenAI 靠 RL 榨干基座模型，Astra 6.1 会更强

Ananth @Ananth7e原推发布于 2026.09.23

在 X 上打开原推

openai are the best at squeezing performance out of the base model through RL. we've already seen this multiple times with previous models. next astra will be the same story here. astra 6.1 will be more useful and way better model than astra-6.

这条是引用着另一条写的：「爆料：Opus 5.5 表现让 OpenAI 措手不及」