Artificial Analysis now labels its Intelligence Index v5.0.1. Its version history places the patch in November 2026, without publishing an exact transition day. We checked the current leaderboard and methodology on November 4.

Regrading, not a new model release

The methodology says the revision improves grading robustness for CritPt and regrades all models on CritPt v2. Our current snapshot changes Gemini 4 Carbon (high) from 59 to 60, Claude Opus 5.5 (max with fallback) from 58 to 59, Sonnet 5.5 (max with fallback) from 54 to 55, and DeepSeek V4.1 Flash 1017 (max) from 32 to 33. These movements follow an evaluation revision and do not establish that the providers changed their models.

All 15 tracked entries were checked against the same current index version. Eleven numerical scores are unchanged. Reasoning effort and fallback settings remain attached to the exact variants; historical articles retain their original dated figures.

Other measurements stay separate

Our Terminal-Bench and Arena measurements are separate datasets. We have not converted vendor demonstrations into independent scores, filled missing model results with sibling values, or treated an index revision as a universal capability ranking.