Compare AI models
Side-by-side benchmark profiles, pricing, and capabilities for up to 4 models. The link carries this exact comparison.
Gemini 4 CarbonGoogle
Claude Opus 5.5Anthropic
GPT-6 AstraOpenAI
Muse WonderMeta
Benchmark profile
Radar chart comparing Gemini 4 Carbon, Claude Opus 5.5, GPT-6 Astra, Muse Wonder across 8 benchmarks. The same scores are in the benchmark table on this page.
Gaps mean no published result, not a zero: Gemini 4 Carbon (Rolling Avg, Data Analysis, HLE); Muse Wonder (SimpleQA, HLE); Claude Opus 5.5 (HLE).
Specs at a glance
Analyzer Index v1.095.7 (5/13)
Family—
Released2026-11-10
Context1.0M
Output limit944k
API features
Input $/M$1.00
Output $/M$5.00
Output speed61 tok/s
Time to first token23.12 s
Cheapest host—
Modalitiestext, vision, audio
Analyzer Index v1.093.8 (6/13)
Family—
Released2026-09-22
Context1.0M
Output limit128k
API features
Input $/M$4.00
Output $/M$20.00
Output speed105 tok/s
Time to first token15.96 s
Cheapest host—
Modalitiestext, vision
Analyzer Index v1.091.3 (7/13)
Family—
Released2026-09-04
Context1.1M
Output limit128k
API features
Input $/M$5.00
Output $/M$25.00
Output speed42 tok/s
Time to first token6.81 s
Cheapest host—
Modalitiestext, vision
Benchmark detail
| Benchmark | Gemini 4 Carbon | Claude Opus 5.5 | GPT-6 Astra | Muse Wonder |
|---|---|---|---|---|
Chatbot Arena Elo preference | ||||
Output Speed performance | ||||
Time to First Token performance | ||||
Rolling Contamination-Controlled Average general | — | |||
Rolling Data Analysis data analysis | — | |||
SimpleQA Verified knowledge | — | |||
OTIS Mock AIME 2024-2025 math | ||||
Humanity's Last Exam reasoning | — | — | — | |
ARC-AGI 2 reasoning | — | |||
Frontier Composite composite | ||||
Output Stability reliability | — | — | ||
Format Adherence reliability | — | — | ||
Recovery Rate reliability | — | — | ||
API Success Rate reliability | ||||
Safety Handling reliability | — | — |