1
Claude Opus 5 MaxAnthropic
Opus 5 Max2 sourcesAug 12, 2026StaleHigh confidence
ToolCenter 58.6
View detailsIndependent model ranking
AI Model Capability Rankings
A traceable model selection view across human preference, reasoning, coding, and open-weight benchmarks.
Formal models
34
Synced sources
4
Pending review
71
Public multi-source snapshots · incompatible raw scores are never addedSnapshot: 8/17/2026
ToolCenter’s composite reference score requires at least two independent sources.
At least two independent public sources
2
GPT-5.6 Sol MaxOpenAI
5.6 Sol Max2 sourcesAug 12, 2026StaleHigh confidence
ToolCenter 58.4
View details3
Claude Opus 4.6Anthropic
4.62 sourcesAug 12, 2026StaleHigh confidence
ToolCenter 58.0
View details4
1.2 xHigh2 sourcesAug 12, 2026StaleHigh confidence
ToolCenter 56.2
View details5
GPT-5.5 HighOpenAI
5.5 High2 sourcesAug 12, 2026StaleHigh confidence
ToolCenter 54.4
View details6
Claude Opus 4.6 ThinkingAnthropic
4.6 Thinking2 sourcesAug 12, 2026StaleHigh confidence
ToolCenter 53.4
View details7
3.8 Max2 sourcesAug 12, 2026StaleHigh confidence
ToolCenter 52.2
View details8
Gemini 3 ProGoogle
3 Pro2 sourcesAug 12, 2026StaleHigh confidence
ToolCenter 51.1
View details9
Gemini 3.7 Flash HighGoogle
3.7 Flash High2 sourcesAug 12, 2026StaleHigh confidence
ToolCenter 50.9
View details10
Claude Opus 4.7 ThinkingAnthropic
4.7 Thinking2 sourcesAug 12, 2026StaleHigh confidence
ToolCenter 50.3
View details11
GPT-5.2 CodexOpenAI
5.2 Codex2 sourcesAug 10, 2026FreshHigh confidence
ToolCenter 26.5
View detailsSources and confidence
Every score keeps its source rank, raw score, source update time, and methodology link. Stale snapshots are labeled instead of being presented as live data.
How the composite reference works
The composite uses normalized percentiles. It is a ToolCenter reference, not an official aggregate score from any benchmark.
Human preference35%
Reasoning25%
Coding / Agent25%
Knowledge & stability15%
ToolCenter’s composite reference score requires at least two independent sources.