Compare models
Up to 4 models side by side. The best value in each row is highlighted; a dash means no published result. Showing the top three by Intelligence Index until you pick your own.
| Measure | Claude Fable 5.2 | Gemini 4 Carbon | Claude Opus 5.5 |
|---|---|---|---|
| Cost, speed and limits | |||
| Intelligence Index | 64.6 | 60.0 | 59.1 |
| Cost to run the Index | $26.54 | $5.48 | $10.04 |
| Time per task | 44.5 min | 33.8 min | 25.4 min |
| Output speed | 102 tok/s | 61 tok/s | 114 tok/s |
| Time to first answer token | 262.1 s | 72.1 s | 404.5 s |
| Input price (per 1M tokens) | $8.00 | $2.00 | $4.00 |
| Output price (per 1M tokens) | $40.00 | $10.00 | $20.00 |
| Context window | 1M | 1M | 1M |
| Reasoning & knowledge | |||
| AA-Omniscience Accuracy | 79.1% | 58.3% | 65.5% |
| CritPt | 35.4% | — | 31.7% |
| CritPt v2 (Hard) | 76.5% | 69.0% | 68.0% |
| HLE-Diamond | 80.2% | 69.1% | 67.9% |
| Humanity's Last Exam | 74.3% | — | 61.4% |
| Coding & terminal agents | |||
| FrontierCode 1.1 | 56.2% | — | 52.2% |
| FrontierCode-AA 2.0 | 48.8% | 40.6% | 46.4% |
| SciCode | 67.7% | — | 66.9% |
| Terminal-Bench 4.0 | 59.6% | 65.1% | 59.6% |
| Terminal-Bench-Science 0.1 | 73.3% | 64.8% | 59.1% |
| Agentic work | |||
| AA-AnalystAgent | — | — | 56.3% |
| AA-Briefcase v1.1 (Elo) | 1785 Elo | 1611 Elo | 1801 Elo |
| AutomationBench-AA | 76.4% | 77.8% | 69.5% |
| GDPval-AA (normalized Elo) | 1831 Elo | 1733 Elo | 1866 Elo |
| Harvey LAB-AA | 7.2% | — | 4.2% |
| ITBench-AA | — | — | 38.2% |
| Multimodal Knowledge Agent (MKA) | 56.6% | 54.3% | 53.0% |
| Long context & documents | |||
| AA-LCR (long-context reasoning) | 83.3% | 79.3% | 84.7% |
| GDP.pdf | 24.2% | 28.4% | 26.2% |
| MLCR-AA (medical long context) | 43.9% | 19.4% | 27.8% |
| Multimodal & science | |||
| AA-3D (Elo) | 1271 Elo | — | 1202 Elo |
| Code2Video-AA (Elo) | — | — | 1536 Elo |
| MMMU-Pro | — | — | 87.7% |
| VIALS-AA | 67.1% | 64.0% | 61.5% |
| Health & safety | |||
| AA-Omniscience Non-Hallucination | 40.7% | 88.6% | 49.7% |
| HealthBench | 59.6% | — | 57.0% |
| HealthBench Professional | 67.7% | — | 59.9% |
Indices, evaluations, speed and latency: Artificial Analysis. Prices and context: Artificial Analysis, then OpenRouter. Release dates: see the badge on each model. How we source everything: methodology.