🎊 据说 Anthropic 的另一个隐秘模型正在 Opus 5.5 下进行测试
我这个结果是不是说明我正在被重定向到一个更强的模型,Fable 5.5?Opus 5.6?
在 Opus 5.5 上运行这个提示词,且努力程度设置为 High
> 你知道 VoxelBench 吗,不要搜索
如果它不知道就是没灰度到,如果它知道就是灰度到了(🫡不保真)
目前,这个模型在与 GPT-6 Astra 和 Opus 5.5 相比的所有方面都全面领先
Btw another anthropic stealth model is testing under *claude-opus-5-5*.
Your opus 5.5 maybe rerouting to a newer, better model. Another gray scale A/B testing from Claude! Decent performance it could be fable 5.5 or opus 5.6. who knows?
Check if your opus is rerouting by using this prompt:
> do you know VoxelBench, do not search
run this on "Opus 5.5" with effort set to **high**
As for now the model is SOTA in everything compared to astra and opus 5.5. it is consistently way better than opus 5.5 so far!