Artificial Analysis, an independent AI analysis firm, released version 1.1 of its Capability Indices on September 14. The update adds what the firm calls stronger domain tuning, blending slices of its core Intelligence Index v4.3 with specialized evaluations built for specific fields.
The distinguishing feature of the Capability Indices is their sourcing: tasks are drawn from O*NET, the U.S. Department of Labor’s occupational database, and mapped to benchmarks meant to represent them. That approach ties model scoring to job-relevant tasks rather than only academic test suites.
Per the firm’s own summary of the release, Claude Fable 5.1, run at maximum effort settings, leads across all six indices in the new version. Artificial Analysis did not publish the individual index names or scores in the material reviewed for this brief, and no runner-up models were disclosed.
The ranking rests on Artificial Analysis’s own methodology and its own summary of the results. Readers evaluating model choice on occupation-relevant tasks should treat this as one benchmarking firm’s framework rather than a settled industry standard, and watch for the full index breakdown before adjusting vendor decisions.
Artificial Analysis, posted on X on September 14, 2026.
