A new scoreboard for AI: Claude Opus 5.5 leads 3 of 9 tests, five other models split the remaining 6

Artificial Analysis released Intelligence Index v5 on Oct. 28. No model leads all nine evaluations.

Claude Opus 5.5 leads three, including the coding test FrontierCode 2.0 (46%) and HLE-Diamond (68%); five other models lead the other six. On Terminal-Bench-Science, GPT-6 Astra (max) leads at 63% with Opus 5.5 (max) at 59%. On HLE-Diamond, GPT-6 Astra is second at 64%.

Sixty percent of the index now rests on private test sets or held-out answers. FrontierCode 2.0, from Cognition, has 132 tasks in real open-source repos and is graded on whether a maintainer would merge the agent's change. The group also launched a Progress Index v1 that tracks capability over time on one scale. The announcement drew 4,000 likes.

Separately, AI YouTuber Matthew Berman wrote that "deepseek v4.1 flash is the best open weights model on the planet right now" (1,800 likes). That claim is unverified and not part of the index results.