Epoch's Capabilities Index puts Claude Fable 5.2 at a record 174, above GPT-6 Astra's 169
Epoch says the score sits above what the pre-Astra trend for reasoning models predicted; it comes from Epoch's own pre-launch testing and may change as more results arrive.
Why it matters
Epoch's ECI combines results from many benchmarks into one scale, so a five-point jump over the previous leader is an independent check on Anthropic's own launch claims. Epoch notes the score beats the earlier trend for reasoning models and links it to the debate over AI contributing to AI research and development, promising more work on that question. The figure is provisional: it uses only Epoch's own benchmarks run before launch. Watch whether the lead holds as other evaluators publish and as rival labs release models.
Key facts
- Claude Fable 5.2 ECI score: 174
- Previous high: GPT-6 Astra at 169
- Above the pre-Astra trend for reasoning models
- Based on Epoch's pre-launch testing with its own benchmarks
- Score may change as more results arrive