Claude Fable 5.1 tops Artificial Analysis index
Anthropic's Claude Fable 5.1 scored 66 on the Artificial Analysis Intelligence Index, the highest score the firm has measured, but each task costs 20% more than on Fable 5 despite a large cut to cache pricing.
Artificial Analysis, which ran pre-release evaluations for Anthropic, reports that Claude Fable 5.1 at max effort scores 66 on its Intelligence Index. That puts it ahead of Claude Opus 5 (63), Claude Fable 5 (62), GPT-5.6 Sol (61) and Grok 4.6 (61). As we reported at launch, Fable 5.1 replaces Fable 5 at the top of Anthropic's lineup.
The 4-point gain over Fable 5 shows up across several benchmarks. Fable 5.1 scores 59.1% on HLE, up from 55.5%. It posts narrow highs on Terminal-Bench v2.1 (91.4%) and SciCode (62.0%), and it adds 9 points on 𝜏³-Banking. Artificial Analysis tested with Anthropic's default safety fallback, which sends flagged requests to Opus 4.8 or Opus 5. The fallback served about 4% of output tokens.
Anthropic cut the cache read price from $1 to $0.25 per million tokens and kept the $10/$50 input/output prices. Even so, Fable 5.1 (max) costs $3.76 per index task, against $3.14 for Fable 5 and $2.34 for Opus 5, because it produces about 1.7x the output tokens of Fable 5. Without the cache cut, a task would cost about $5.16. At xhigh effort, the model scores 65 for $2.72 per task.
On agentic knowledge-work tests, Fable 5.1 leads GDPval-AA v2 with 1,853 Elo against 1,824 for Opus 5, though the confidence intervals overlap. On AA-Briefcase the two are effectively tied. Fable 5.1 also answers more AA-Omniscience questions and reaches a record 67.2% accuracy, but it also hallucinates more often, so its Omniscience Index score is level with Fable 5's.