Claude Fable 5.2 takes #1 on Epoch's Capabilities Index at 174, five points clear of GPT-6 Astra

Claude Fable 5.2 is the new #1 on the Epoch Capabilities Index (ECI), Epoch AI's composite of more than 50 benchmarks. Epoch's public score file, updated November 9, puts it at 174.2, five points ahead of GPT-6 Astra at 169.2. It scores 255 models.

THE TOP FIVE

• Claude Fable 5.2: 174.2 (credible interval 170.1 to 178.8)
• GPT-6 Astra: 169.2
• GPT-6.1 Sol: 168.9
• Claude Opus 5.5: 168.7
• GPT-6.1 Terra: 168.1
• Claude Fable 5.1, the model it replaces, is sixth at 167.0, so the point release added 7.2 points.

WHAT A FIVE-POINT LEAD MEANS

Epoch says absolute ECI values mean nothing on their own; only comparisons between models do. It also says that when the index launched, a 5-point gain roughly matched a doubling of METR's time horizon, the length of task a model can finish. Fable 5.2's credible interval still overlaps GPT-6 Astra's at the low end (170.1 against Astra's upper bound of 172.5), so the lead is large but not outside the error bars.

ALSO FIRST ON ARTIFICIAL ANALYSIS

Artificial Analysis scores Fable 5.2 (Max) at 65 on its Intelligence Index v5.0.1, the top score on the index. Its page also flags the cost of that score: 275M output tokens to run the index against a median of 72.8M, 45 minutes per task, and $26.54 per task to evaluate.

WHAT WE DO NOT KNOW

How far the number will move. Fable 5.2's interval is wider than those of the models behind it (8.7 points, against 6.3 for GPT-6 Astra), so the score may shift as Epoch adds benchmark results. LuminaBench on X was first to point out the new ranking; the figures above are read from Epoch's and Artificial Analysis's own pages.