Artificial Analysis overhauls its Intelligence Index
Artificial Analysis has released version 5 of its Intelligence Index, which now relies mostly on private test sets and shows the gap between open-weights and frontier models widening from 3 points to 12.
Artificial Analysis announced Intelligence Index v5, its biggest revision since v4 launched in January. Private test sets or held-out answers now make up 60% of the composite score. The firm also launched a Progress Index v1, which tracks model capability over time on one continuous scale.
v5 adds two benchmarks. The first is Cognition's FrontierCode 2.0, 132 private tasks in real open-source repositories, graded on whether a maintainer would merge the agent's change. The second is Terminal-Bench-Science, 70 agentic research tasks run in a terminal. HLE-Diamond replaces Humanity's Last Exam, and AA-Omniscience and CritPt were also upgraded. GDPval-AA, SciCode and AA-LCR were dropped. Knowledge work, coding, science and general ability each count for 25% of the score.
Claude Opus 5.5 (max) leads with 58, ahead of GPT-6 Astra (57) and GPT-6.1 Sol (55). Anthropic and OpenAI hold the top five places, and Gemini 4 Argon is next at 50. No single model leads all nine evaluations. Opus 5.5 tops three of them, and GPT-6 Astra leads Terminal-Bench-Science (63%) and the CritPt hard subset (72%).
The top open-weights model is Kimi K3.1 Pro at 46, but its weights have not yet been released. Among models with released weights, DeepSeek V4.1 Flash 1017 and GLM-5.3 lead at 32. That is 26 points behind the leader and below Claude Haiku 5.5 (36). Some of the wider gap reflects the new benchmark mix rather than a change in what the models can do.
Among the ten highest-scoring models, cost per task varies more than 13x, from $0.96 for GPT-6.1 Terra to $13.04 for Claude Fable 5.1.