58index points
060 index points
Other scores & settings
- Coding
- 59.6%
- Reasoning
- 61.4%
- Work
- 1846Elo
Adaptive reasoning · max effort · default fallback
Compare AI model performanceAI benchmarks
Compare models for coding, reasoning, and everyday work.
23 models · 4 benchmarksSnapshot checked Source: Artificial Analysis
Compare model pricing & context windows Explore more AI statistics & research websites Overall capability
Artificial Analysis Intelligence Index v4.3.2 · Higher is better · index points
23 of 23 models · Ranked within this selection
- 1Claude Opus 5.5Anthropic58
- 2Claude Sonnet 5.5Anthropic56
- 3TiedClaude Fable 5.1Anthropic53
- 3TiedGPT-6 AstraOpenAI53
- 4GPT-6.1 SolOpenAI52
Axis: 0–60 index points. Scale stays fixed when filtering. Showing up to five models; more tied models may appear in All models.
56index points
060 index points
Other scores & settings
- Coding
- 63.6%
- Reasoning
- 55.0%
- Work
- 1844Elo
Adaptive reasoning · max effort · default fallback
53index points
060 index points
Other scores & settings
- Coding
- 55.1%
- Reasoning
- 58.7%
- Work
- 1721Elo
Adaptive reasoning · xhigh effort · default fallback
53index points
060 index points
Other scores & settings
- Coding
- 59.1%
- Reasoning
- 54.7%
- Work
- 1542Elo
Reasoning · max effort
52index points
060 index points
Other scores & settings
- Coding
- 56.1%
- Reasoning
- 52.9%
- Work
- 1575Elo
Max reasoning effort
Scales stay fixed when filtering. Colour shading compares scores within each benchmark across the full snapshot. Missing scores are unshaded.
How this benchmark works
Artificial Analysis Intelligence Index v4.3.2
A composite of ten evaluations spanning reasoning, coding, and professional tasks. Index points are not a percentage of tasks passed.
View results at Artificial AnalysisRanks recalculate for the selected benchmark and filters. Equal displayed scores share a rank with no gaps (1, 1, 2). “Not available” means no verified score in this snapshot; 0.0% is a reported score. Scores do not establish statistical significance.
Sources, settings & methodology
A dated snapshot
These scores were checked on 1 October 2026 and are updated manually. This is the source-check date; individual test dates were not available. Source leaderboards may have changed since. This selection covers major model families, not every model or configuration.
Settings change the result
Scores use the model configurations shown in each row. Reasoning effort, fallback models, tools, and evaluation setup affect results. Use the model source links for full details. Rankings use the displayed precision and do not establish statistical significance.
Compare within each benchmark
All scores come from Artificial Analysis. Overall uses Intelligence Index v4.3.2; coding uses Terminal-Bench v4.0; work uses GDPval-AA v2.1. We do not mix earlier versions or calculate a new combined score. Benchmark performance is a starting point for testing your own tasks.