Top stories

Artificial Analysis released Intelligence Index v5 on October 27, 2026, adding two benchmarks and revising three others so that 60% of the composite model-capability score now relies on private test sets or held-out answers. The index measures knowledge work, coding, science and general capabilities at 25% each. New evaluations include Cognition’s 132-task FrontierCode 2.0 coding benchmark and Terminal-Bench-Science, which contains 70 terminal-based scientific research tasks. Claude Opus 5.5 at maximum effort leads the overall index with 58, followed by GPT-6 Astra at 57 and GPT-6.1 Sol at 55; Gemini 4 Argon scores 50. Claude Opus 5.5 leads FrontierCode 2.0 at 46% and HLE-Diamond at 68%, while GPT-6 Astra leads Terminal-Bench-Science at 63% and the CritPt v2 hard subset at 72%. No model led all nine evaluations. Kimi K3.1 Pro, whose weights have not yet been released, is the highest-scoring open-weights model at 46, 12 points behind Claude Opus 5.5. Among models with released weights, DeepSeek V4.1 Flash 1017 and GLM-5.3 lead at 32. Artificial Analysis also launched Progress Index v1, a continuous capability scale covering language models released since November 2022.