Claude Fable 5.2 在 Epoch Capabilities Index（ECI）上以 174 分刷新纪录，超过 GPT-6 Astra 此前的 169 分。
Internal models being between 175 and 180 makes a lot of sense now that we know Anthropic's last gen model scores 174
Still, this was surprising as I thought Fable 5.2 would score 173 (purely based on AECI and vibes). Despite that I think the new Astra will score higher and is the more interesting launch, as this is basically the first generational upgrade of looping models.
It will tell us a lot about whether OpenAI is seriously concerned about CoT monitorability (and safety in general) or whether they are just chasing frontier performance.
and I think it's obvious that safety is still second priority at both Anthropic and OpenAI. Capabilities are most important and labs will continue pushing. (especially next year they will be forced to. chinese loopmaxxers are coming)
I hope OpenAI gives us more insight into loopers, their scaling and safety research. (especially their plans on monitoring them effectively)
Claude Fable 5.2 sets a new record on the Epoch Capabilities Index (ECI) at 174. The previous high was GPT-6 Astra’s 169. Fable 5.2’s score is above what the pre-Astra trend for reasoning models would have predicted. This is notable given recent discussion of AI contributions to AI R&D, and we hope to release more work in this area soon. This is based on testing we did prior to launch using only our own benchmarks. The ECI is continually updated as more benchmark results become available, and may change in the future.在 X 查看被引用的帖子
来源：Lisan al Gaib · x.com