The Pulse

Vals AI ranks Claude Fable 5.2 first on its Vals Index

Vals AI ranked Anthropic’s Claude Fable 5.2 first on its Vals Index, with a score of 71.31%. The model also led four index components, but placed lower on HLAB-Vals-1.1 and Tax Agent Bench.

AI.info Team · Written with AI · fact-checked · 2026-11-10

Vals AI ranked Anthropic’s Claude Fable 5.2 first on its Vals Index, with a score of 71.31%, in an evaluation published Nov. 9. The result put Fable ahead of Claude Sonnet 5.5, at 70.85%, and Gemini 4 Carbon, at 70.45%, according to the benchmark company.

Results across benchmarks

The company reported Fable at No. 1 among 36 models on the index, at a listed cost of $17.94 per test. Vals AI says its benchmark program tests models on economically valuable work, including finance and software tasks, as well as frontier-risk areas such as cybersecurity.

Fable ranked first on four Vals Index components: Finance Agent v2.1, with 69.47%; Vibe Code Bench, with 93.77%; and EMB v1.1, with 77.38%. It tied with Claude Opus 5.5 for first on Terminal-Bench 4.0, with both models scoring 65.15%, Vals AI said.

The report also placed Fable first among 66 models on ProgramBench, where it scored 19.00%, ahead of Opus 5.5 at 18.50%. On SWE-bench Multimodal, it ranked second among 22 models with 62.29%, behind Opus 5.5 at 70.21%.

Science scores and weaker areas

Vals AI reported that Fable led 31 models on BioMysteryBench, scoring 83.70%. It ranked second among 24 models on the Vals Science Index at 63.74%, behind GPT-6 Astra at 65.09%.

The results were lower on two other measures named in the report. Fable scored 5.00% on HLAB-Vals-1.1, tied for 29th among 37 models, and 32.09% on Tax Agent Bench, ranking 27th among 84, according to Vals AI.

Evaluation setup

Vals AI described Claude Fable 5.2 as a reasoning model with a 1-million-token context window. The company listed prices of $8.00 per million input tokens and $40.00 per million output tokens, and said it evaluated the model at High effort.

The report covers the Vals Index v2.2, its components and the Vals Science Index. Vals AI says it conducts its own evaluations and creates many of its benchmarks in-house.