Artificial Analysis released a new version of its AI Intelligence Index on October 28th (Beijing time), showing significant changes in model rankings. In the previous version, Claude Fable 5.1 and GPT-6 Astra tied for first place with 53 points; in the new version, Claude Opus 5.5 tops the list with 58 points, GPT-6 Astra comes in second with 57 points, GPT-6.1 Sol rises to third with 55 points, and Gemini 4 Argon follows with 50 points. In the sub-categories, Opus 5.5 achieved 46% success in FrontierCode 2.0 and 68% in HLE-Diamond; Astra leads in scientific research tasks (63%) and physics reasoning (72%); Kimi K3.1 Pro leads Terminal-Bench 4.0 with 67%, Gemini 4 Argon leads in automated workflows with 78%, and Muse Wonder Full achieves a 33% pass rate in the GDP.pdf document task. The gap between the open-weight model and the top-ranked model has widened from 3 points to 12 points; DeepSeek V4.1 Flash and GLM-5.3, which have released weights, both scored 32 points.
