Vals AI LiveCodeBench: leaderboard
Metric: Accuracy (%). Source: www.vals.ai. 143 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Fable 5.1 (Max) | 90.52 |
| 2 | Claude Fable 5 (Max) | 89.78 |
| 3 | Gemini 3.8 Flash (High) | 89.48 |
| 4 | Claude Opus 5 (Max) | 89.03 |
| 5 | Gemini 3.7 Flash (High) | 88.65 |
| 6 | Gemini 3.1 Pro (Preview) (High) | 88.48 |
| 7 | Grok 4.6 (High) | 88.22 |
| 8 | Gemini 3.6 Flash (High) | 88.08 |
| 9 | GPT-5.2 Codex | 87.99 |
| 10 | Qwen 3.8 Max | 87.85 |
| 11 | Claude Opus 4.8 (Max) | 87.82 |
| 12 | Gemini 3.5 Flash (High) | 87.6 |
| 13 | DeepSeek V4 Pro (0813) (Max) | 87.53 |
| 14 | DeepSeek V4 Pro (Max) | 87.48 |
| 15 | Grok 4.5 (High) | 87.35 |
Interactive version: theaggregate.ai/benchmark?slug=vals-ai-livecodebench · How It Works · Data refreshed daily, snapshot 2026-09-05.