LiveBench: leaderboard
Continuously updated benchmark measuring many capabilities. Sort models by weighted score or sub-task scores. Tests math, coding, reasoning, language, instruction following, and data analysis.
Metric: Score. Source: livebench.ai. Status: saturation imminent. 54 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5.1 (Max) | 92.97 |
| 2 | Claude Fable 5 (Max) | 92.84 |
| 3 | Claude Opus 5 (Max) | 92.8 |
| 4 | Claude Opus 4.8 (Max) | 92.23 |
| 5 | GPT-6 (Max) | 92.17 |
| 6 | Claude Opus 4.6 (Thinking, High) | 92.17 |
| 7 | Gemini 3.8 Flash (High) | 92.16 |
| 8 | Gemini 3.1 Pro (Preview) (High) | 92.1 |
| 9 | Gemini 3.5 Flash (High) | 91.88 |
| 10 | Gemini 3.7 Flash (High) | 91.79 |
| 11 | Muse Spark 1.3 (xHigh) | 91.74 |
| 12 | GPT-5.6 Sol (Max) | 91.69 |
| 13 | Kimi K3 | 91.65 |
| 14 | GPT-5.6 Terra (Max) | 91.53 |
| 15 | GPT-5.4 (xHigh) | 91.47 |
Interactive version: theaggregate.ai/benchmark?slug=livebench · How It Works · Data refreshed daily, snapshot 2026-09-05.