註冊並分享邀請連結,可獲得影片播放與邀請獎勵。

Haider.
@haider1
open intelligence for a decentralized future
加入 November 2021
4.4K 正在關注    72.8K 粉絲
i wouldn't call fable 5.2 a benchmark-maxed model but it's not efficient than fable 5.1 as anthropic claims when the cost per task has doubled from 5.1 to 5.2 it feels more like a model with better post-training than fable 5.1 but only a smaller improvement over opus 5.5
顯示更多