Artificial Analysis Intelligence Index v5: Claude Opus 5.5 Leads at 58, Open-Weights Gap Widens to 12
Artificial Analysis released Intelligence Index v5 on October 27, 2026, its biggest methodology overhaul since v4 launched in January. Sixty percent of the index now runs on private test sets or held-out answers, up from a minority share in prior versions.
What Changed
Two new benchmarks enter the composite:
FrontierCode 2.0 — Cognition’s private coding evaluation, launched the same day. The benchmark contains 132 tasks drawn from real open-source repositories. Each task is graded on whether a repository maintainer would accept the agent’s pull request. Failing any blocking check scores zero. Tasks remain private to prevent contamination.
Terminal-Bench-Science — 70 agentic scientific research tasks run in a terminal environment. Experts curate and review each task. A task passes only when every test passes.
Three existing benchmarks were upgraded: HLE-Diamond (a 1,000-question refinement of Humanity’s Last Exam) replaces the original HLE; AA-Omniscience was updated to v1.1 with contaminated and public-set questions removed; CritPt v2 now runs on a private 40-question hard subset.
Three benchmarks were removed: GDPval-AA v2.1, SciCode, and AA-LCR v1.1.
The index weights are equal across four categories: Knowledge Work (25%), Coding (25%), Science (25%), and General (25%).
Leaderboard
| Model | Index v5 Score |
|---|---|
| Claude Opus 5.5 (max) | 58 |
| GPT-6 Astra (max) | 57 |
| GPT-6.1 Sol (max) | 55 |
| Gemini 4 Argon (high) | 50 |
| Kimi K3.1 Pro (weights unreleased) | 46 |
| DeepSeek V4.1 Flash 1017 (max) | 32 |
| GLM-5.3 (max) | 32 |
Anthropic and OpenAI hold the top five slots. Gemini 4 Argon is the first non-Anthropic, non-OpenAI model in the top four.
No single model leads all nine evaluations. Claude Opus 5.5 leads FrontierCode 2.0 (46%), HLE-Diamond (68%), and AA-Omniscience v1.1 (48). GPT-6 Astra leads Terminal-Bench-Science (63%) and CritPt v2 hard subset (72%). Claude Sonnet 5.5 leads AA-Briefcase at 1,812 Elo; Kimi K3.1 Pro leads Terminal-Bench 4.0 at 67%; Gemini 4 Argon leads AutomationBench-AA at 78%; Muse Wonder Full leads GDP.pdf at 33%.
New Benchmark Scores in Detail
FrontierCode 2.0: Claude Opus 5.5 (max) leads at 46%. GPT-6 Astra (max), GPT-6.1 Terra (max), and Claude Sonnet 5.5 (max) all score 42%. DeepSeek V4.1 Flash 1017 is the highest open-weights model at 28%, 18 points behind the leader.
Terminal-Bench-Science: GPT-6 Astra (max) leads at 63%, with Claude Opus 5.5 (max) at 59% and GPT-6.1 Sol (max) at 58%. Scores drop steeply after the top three: Kimi K3.1 Pro (weights unreleased) scores 34%, and DeepSeek V4.1 Flash 1017 scores 13%. Most open-weights models come in under 10%.
HLE-Diamond: Claude Opus 5.5 (max) leads at 68%. GPT-6 Astra (max) is at 64%. Claude Fable 5.1 (max) scores 60%.
Open-Weights Gap Widens Sharply
The gap between the best open-weights model and the leader widened from 3 points in v4 to 12 points in v5. The jump is concentrated in HLE-Diamond and Terminal-Bench-Science, where the difference between Kimi K3.1 Pro and Claude Opus 5.5 reaches 31 and 25 points, respectively.
Among models with weights actually released, DeepSeek V4.1 Flash 1017 and GLM-5.3 lead open-weights at 32. Both score below Claude Haiku 5.5 (max, 36). The released-weights deficit versus the frontier is 26 points.
Cost Varies by More Than 13x
Among the 10 highest-scoring models, cost per Intelligence Index task ranges from $0.96 (GPT-6.1 Terra, max) to $13.04 (Claude Fable 5.1, max). GPT-6.1 Sol (max) scores 55 — within 4 points of Claude Opus 5.5 — at $1.27 per task, roughly one-eighth the cost of Claude Opus 5.5 at $10.04.
GPT-6 Astra (max) scores 57 at $6.24 per task.
Effort-level spread also widened compared to v4. Claude Opus 5.5 spans from low effort (score unspecified) to 58 at max, with cost per task rising from $0.98 to $10.04 at max. Claude Sonnet 5.5 spans 29 points from low to max effort, up from 20 points in v4.3. GPT-6.1 Sol’s spread doubled from 10 to 20 points.
Progress Index v1
Alongside v5, AA launched the Artificial Analysis Progress Index v1, which places models released since November 2022 on a single capability scale, designed to make cross-generation comparisons consistent. The Progress Index is separate from the Intelligence Index and is intended for longitudinal tracking rather than current-model ranking.