Artificial Analysis Intelligence Index v5: Claude Opus 5.5 Leads as Open Weights Gap Widens to 12 Points

• Artificial Analysis Intelligence Index v5 introduces two new benchmarks (FrontierCode 2.0, Terminal-Bench-Science) and upgrades three existing ones, shifting 60% of the index to private datasets or held-out answers.
• Claude Opus 5.5 (max) leads the index with a score of 58, ahead of GPT-6 Astra (max, 57) and GPT-6.1 Sol (max, 55), while the gap between the top model and the best open-weights model widens from 3 to 12 points.
• Cost per task varies significantly among top models, ranging from $0.96 for GPT-6.1 Terra (max) to $13.04 for Claude Fable 5.1 (max), with GPT-6.1 Sol (max) offering near-frontier performance at a fraction of the cost.
• Effort settings impact scores more in v5, with Claude Opus 5.5 scoring 35 at low effort versus 58 at max effort, and median output tokens per task rising from 35k to 59k.

Benchmark Changes

• FrontierCode 2.0: 132 private real-world coding tasks graded on maintainer merge acceptance; Claude Opus 5.5 leads at 46%.
• Terminal-Bench-Science: 70 agentic scientific research tasks; GPT-6 Astra leads at 63%, with open-weights models scoring under 10%.
• HLE-Diamond: Replaces Humanity's Last Exam with a refined 1,000-question set; Claude Opus 5.5 leads at 68%.

New Progress Index

• Artificial Analysis Progress Index v1 tracks model capability over time on a continuous scale for models released since November 2022.
• Designed to compare underlying capability changes across years, complementing the Intelligence Index which measures current frontier performance.