Artificial Analysis releases Intelligence Index v5, with Claude Opus 5.5 on top

Artificial Analysis published version 5 of its Intelligence Index, a composite of model capability, with two new benchmarks and three upgrades. It says that for the first time most of the index is built on private datasets. It also launched a Progress Index v1 that places language models released since November 2022 on one capability scale.

What the benchmark site reports (its own measurements):

• New: FrontierCode 2.0, Cognition's private coding benchmark of 132 tasks in real open-source repositories, graded on whether a maintainer would merge the change; and Terminal-Bench-Science, where GPT-6 Astra (max) leads at 63%, followed by Claude Opus 5.5 (max) at 59%.
• Upgraded: HLE-Diamond replaces Humanity's Last Exam (Claude Opus 5.5 (max) leads at 68%, GPT-6 Astra (max) 64%), AA-Omniscience v1.1 is a cleaned version of its knowledge and hallucination test, and CritPt v2 is scored on a private 40-question hard subset.
• Weights: Knowledge Work 25%, Coding 25% (FrontierCode 2.0 15%, Terminal-Bench 4.0 10%), Science 25%.
• No single model leads all nine evaluations. Claude Opus 5.5 leads three, and five other models lead the remaining six.
• Open-weights models fall further behind: the leader, Kimi K3.1 Pro (weights not yet released), scores 46, 12 points behind the top, against a 3-point gap in v4.3. Of models with released weights, DeepSeek V4.1 Flash 1017 (max) and GLM-5.3 (max) score highest at 32.
• Cost per index task: Claude Opus 5.5 (max) $10.04, GPT-6 Astra (max) $6.24, GPT-6.1 Sol (max) $1.27. Sol scores within 4 points of Opus 5.5 at about an eighth of the cost.
• Effort matters more: Opus 5.5 scores 35 at low effort and 58 at max, at 10 times the cost per task ($0.98 against $10.04).
• Models use more output tokens per task than in v4.3: the median rose from 35k to 59k.

A relay on X reports GPT-6.1 Terra (max) at 51 on the new index.