Intelligence Index v5 puts Claude Opus 5.5 on top and widens open-weight gap

Artificial Analysis launched the Intelligence Index v5 in late October, its composite ranking of AI models — a benchmark, that is, a standardized battery of tests to compare capabilities — and the top spot changed hands. Claude Opus 5.5, at maximum reasoning effort, took the lead with 58 points in Artificial Analysis' Intelligence Index v5, ahead of OpenAI's GPT-6 Astra with 57 points and GPT-6.1 Sol with 55 points.

The frontier race remains tight, with little separating the leaders even under a tougher test. Next come Claude Sonnet 5.5, Gemini 4 Argon and Meta's Muse Wonder Full, plus Moonshot AI's Kimi K3.1 Pro, as the score chart shows.

A tougher, more closed test

The bigger change is in the yardstick, not just the podium. For the first time, most of the score comes from private tests the labs have never seen before, according to Artificial Analysis itself. The release brings two new evaluations and three updated benchmarks, in the biggest overhaul since the index was created, plus an update to the general knowledge section to replace saturated tests.

The open side of the industry felt the new yardstick most. The gap between the best closed model and the best open-weight model jumped from 3 to 12 points under the new methodology, according to Artificial Analysis. Moonshot AI's Kimi K3.1 Pro leads open models with 46 points, well behind Opus 5.5's 58 points.

That reverses the catch-up trend seen months earlier. In the old v4.3, Kimi K3 had closed to within 6 points of the leader, after open models halved the gap over the previous year. The K3.1 Pro is a mixture-of-experts model with 2.8 trillion total parameters and 104 billion active at a time, with a context window — the amount of text it can process at once — of 1 million tokens.

Costly top, cheap alternative

Being the smartest doesn't mean being the most economical to run in agents and apps. Artificial Analysis' cost-per-task chart compares Claude Opus 5.5 with GPT-6.1 Sol, GPT-6 Astra, Gemini 4 Argon and Kimi K3.1 Pro, and shows a wide gap between the leader and cheaper alternatives.

Meta's Muse Wonder Full shows how the ranking measures the yardstick as much as the model. Two weeks before v5, it had scored 56 points on the previous version, essentially tied with Astra in mid-October. It didn't get worse: the test changed and put it on a different competitive footing.

For enterprises, the index carries weight because of its independence: Artificial Analysis is not a lab and doesn't sell models, so procurement teams use the score to justify large inference contracts. Alongside v5, it launched Progress Index v1, which places all major language models since November 2022 on the same capability scale to show evolution over time.