Coding benchmark
LiveCodeBench (Vals) leaderboard
LiveCodeBench, Vals AI run. Every model the catalog carries a published LiveCodeBench (Vals) value for, ranked by that value.
Vals AI’s independent implementation of the LiveCodeBench code-generation benchmark, run under one fixed harness across the models it tracks.
LiveCodeBench (Vals) ranking
48 models with a published LiveCodeBench (Vals) value, ordered by that value, highest first. Models the source has not scored on this benchmark are not listed — they are unmeasured, not last.
| Rank | Model | Provider | Pass@1 accuracy |
|---|---|---|---|
| 1 |  Claude Fable 5.1 | Anthropic | 90.5 |
| 2 | Claude Fable 5 | Anthropic | 89.8 |
| 3 |  Gemini 3.8 Flash | 89.5 | |
| 4 | Claude Opus 5 | Anthropic | 89.0 |
| 5 | Gemini 3.7 Flash | 88.7 | |
| 6 | Gemini 3.1 Pro | 88.5 | |
| 7 |  Grok 4.6 | xAI | 88.2 |
| 8 | Gemini 3.6 Flash | 88.1 | |
| 9 |  Qwen3.8 Max | Alibaba | 87.9 |
| 10 | Claude Opus 4.8 | Anthropic | 87.8 |
| 11 | Gemini 3.5 Flash | 87.6 | |
| 12 |  DeepSeek V4 Pro 0813 | DeepSeek | 87.5 |
| 13 | Grok 4.5 | xAI | 87.4 |
| 14 | DeepSeek V4 Flash 0731 | DeepSeek | 87.3 |
| 15 |  Kimi K3 | Moonshot AI | 87.2 |
| 16 | Qwen3.7 Max | Alibaba | 87.1 |
| 17 | Kimi K2.6 | Moonshot AI | 86.8 |
| 18 |  Nemotron 3 Ultra | NVIDIA | 86.0 |
| 18 | Qwen3.6 Plus | Alibaba | 86.0 |
| 20 |  GPT-5.6 Terra | OpenAI | 85.9 |
| 20 | TMInkling-Small | Thinking Machines Lab | 85.9 |
| 20 |  Muse Spark 1.1 | Meta | 85.9 |
| 23 | Gemini 3 Flash | 85.6 | |
| 24 | TMInkling | Thinking Machines Lab | 85.5 |
| 25 | GPT-5.5 | OpenAI | 85.3 |
| 26 | Claude Opus 4.7 | Anthropic | 85.1 |
| 27 | Grok 4.3 | xAI | 84.5 |
| 28 | Grok 4.20 | xAI | 84.3 |
| 29 | GPT-5.4 nano | OpenAI | 84.0 |
| 29 |  Ling 3.0 Flash | InclusionAI | 84.0 |
| 29 | Qwen3.8-27B | Alibaba | 84.0 |
| 32 | GPT-5.6 Sol | OpenAI | 82.6 |
| 33 | Claude Sonnet 5 | Anthropic | 82.4 |
| 34 |  MiniMax M3 | MiniMax | 82.2 |
| 35 | Claude Sonnet 4.6 | Anthropic | 82.1 |
| 36 | GPT-5.4 mini | OpenAI | 81.5 |
| 36 |  MiMo-V2.5 | Xiaomi | 81.5 |
| 38 |  GLM-5.1 | Z.AI | 81.4 |
| 38 | MiMo-V2.5-Pro | Xiaomi | 81.4 |
| 40 | GLM-5.3 | Z.AI | 80.5 |
| 40 | GLM-5.3-Flash | Z.AI | 80.5 |
| 42 | Gemini 3.1 Flash-Lite | 80.1 | |
| 43 | MiniMax M2.7 | MiniMax | 79.9 |
| 44 | Gemini 3.5 Flash-Lite | 79.0 | |
| 45 | GLM-5.2 | Z.AI | 69.5 |
| 46 |  Laguna M.1 | Poolside | 68.1 |
| 47 | Laguna XS.2 | Poolside | 67.8 |
| 48 | Claude Haiku 4.5 | Anthropic | 41.2 |
Evidence key: Observed
Rows are ordered by the value Vals AI LiveCodeBench, Vals AI run leaderboard published, highest first. The source does not state whether a higher value is the better result, so this page does not either: for a benchmark that measures a rate of failure, read the table from the bottom.