Claude Opus 5.5 Tops All Six AA Indexes

This is a strong sweep, but it reflects Artificial Analysis’s chosen weights and evaluations—not universal proof that Opus is best for every workflow.

• –v2.0 aligns with Intelligence Index v5, adding or upgrading private and harder evaluations while removing GDPval-AA from the composite.
• –The shift toward tool use and factual reliability makes the indexes more relevant to production work than pure knowledge tests.
• –Kimi K3.1 Pro leads among open-weight models, though its weights remain unreleased.
• –Rankings are domain-specific and sensitive to benchmark composition; developers should inspect component scores before switching models.
• –Opus’s lead may still lose on cost, latency, or availability, so benchmark dominance does not automatically equal best ROI. ([Artificial Analysis methodology](https://artificialanalysis.ai/methodology/capability-indices), [indexes](https://artificialanalysis.ai/indexes))