Benchmarks · composite
Epoch Capabilities Index
Epoch AI's Capabilities Index: a single score per model fitted across many benchmarks (a Bayesian item-response model since Nov 4, 2026), so scores are comparable across models tested on different sets. Claude 3.5 Sonnet is set to 130 and GPT-5 to 150. Intervals are 90% credible intervals. Epoch keeps each model's best score per benchmark across settings, so a row is the model's best measured setting, not one reasoning effort. Data: Epoch AI, CC BY.
Epoch refitted the Capabilities Index with a Bayesian model. Models above 150 moved up about 0.7 on average, and the 76 models below 125 fell about 14. Values before Nov 4, 2026 are not comparable with later ones. Their method
Each row is the model’s best measured setting, not one reasoning effort: the publisher keeps each model’s best score across settings.
171 models · 171 results · data as of 2026-11-06 · epoch.ai ↗
Leaderboard
All results
171 rows| # | Model | Effort | Epoch Capabilities Index |
|---|---|---|---|
| 1 | GPT-6 Astra | best measured setting | 169.1 ±3.1 |
| 2 | GPT-6.1 Sol | best measured setting | 168.9 ±3.1 |
| 3 | Opus 5.5 | best measured setting | 168.4 ±3.0 |
| 4 | GPT-6.1 Terra | best measured setting | 167.8 ±3.1 |
| 5 | Fable 5.1 | best measured setting | 166.9 ±2.8 |
| 6 | Gemini 4 Argon | best measured setting | 166.4 ±2.8 |
| 7 | Sonnet 5.5 | best measured setting | 166.2 ±2.9 |
| 8 | Opus 5 | best measured setting | 164.4 ±2.6 |
| 9 | Fable 5 | best measured setting | 164.2 ±2.5 |
| 10 | GPT-6 Sol | best measured setting | 164.1 ±2.6 |
| 11 | GPT-5.6 Sol | best measured setting | 163.0 ±2.4 |
| 12 | GPT-5.5 Pro | best measured setting | 162.7 ±2.8 |
| 13 | Muse Wonder (Max) | best measured setting | 162.0 ±2.5 |
| 14 | GPT-5.6 Terra | best measured setting | 160.9 ±2.4 |
| 15 | Haiku 5.5 | best measured setting | 160.2 ±2.3 |
| 16 | GPT-5.4 Pro | best measured setting | 160.2 ±2.6 |
| 17 | GPT-5.5 | best measured setting | 160.0 ±2.1 |
| 18 | Opus 4.8 | best measured setting | 159.9 ±2.1 |
| 19 | DeepSeek V4.1 Flash 1017 | best measured setting | 159.3 ±2.4 |
| 20 | Gemini 3.8 Flash | best measured setting | 158.6 ±2.1 |
| 21 | Kimi K3 | best measured setting | 158.5 ±2.0 |
| 22 | Gemini 3.7 Flash | best measured setting | 158.4 ±2.1 |
| 23 | Muse Spark 1.3 | best measured setting | 158.0 ±2.1 |
| 24 | Grok 4.6 | best measured setting | 157.9 ±2.0 |
| 25 | GPT-5.4 | best measured setting | 157.8 ±1.9 |
| 26 | Opus 4.7 | best measured setting | 157.6 ±1.9 |
| 27 | Qwen 3.8 Max | best measured setting | 157.4 ±2.2 |
| 28 | GPT-5.2 Pro | best measured setting | 157.2 ±2.3 |
| 29 | GLM 5.3 | best measured setting | 156.9 ±2.2 |
| 30 | GPT-5.6 Luna | best measured setting | 156.8 ±2.0 |