Artificial Analysis rebuilds its Intelligence Index, and the gap between the best closed and open-weight models grows from 3 points to 12

On 27 October the benchmarking firm Artificial Analysis released version 5 of its Intelligence Index, the composite score many developers use to compare models. It added FrontierCode 2.0 and Terminal-Bench-Science, replaced Humanity's Last Exam with HLE-Diamond, updated two other tests, and removed GDPval-AA v2.1, SciCode and AA-LCR v1.1.

Under version 5, Claude Opus 5.5 (max) leads with 58, ahead of GPT-6 Astra (max) on 57 and GPT-6.1 Sol (max) on 55. Gemini 4 Argon (high) scores 50. Moonshot's Kimi K3.1 Pro, whose weights are not yet released, scores 46. Among models whose weights are out, DeepSeek V4.1 Flash 1017 and GLM-5.3 tie at 32.

The firm says: “The open weights gap widens from 3 points to 12”. Much of that comes from the harder new science and terminal tests, so the change partly reflects which tests were chosen, not only what the models can do.

The index matters because labs quote it in launches and buyers use it to pick models. A wider gap is a talking point for closed-model sellers. It also gives a measure to watch when Moonshot releases Kimi K3.1 Pro's weights, which it has promised by 4 November.

• Confirmed Version 5 scores: Claude Opus 5.5 (max) 58, GPT-6 Astra (max) 57, GPT-6.1 Sol (max) 55, Gemini 4 Argon (high) 50. Artificial Analysis
• Confirmed The gap between the best closed and the best open-weight models widens from 3 points to 12; DeepSeek V4.1 Flash 1017 and GLM-5.3 tie at 32. Artificial Analysis