LiveCodeBench (Vals AI public dataset)
Code · 2024-03-12
Vals AI LiveCodeBench public-dataset evaluation surface with its own sampling, max-output-token, temperature, cost, and latency settings. Keep separate from official LiveCodeBench and Artificial Analysis rows.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5.1 | 90.5 |
| Claude Fable 5 | 89.8 |
| Gemini 3.8 Flash | 89.5 |
| Claude Opus 5 | 89.0 |
| Gemini 3.7 Flash | 88.7 |
| Gemini 3.1 Pro Preview | 88.5 |
| Grok 4.6 | 88.2 |
| Gemini 3.6 Flash | 88.1 |
| GPT-5.2-Codex | 88.0 |
| Qwen3.8-Max | 87.9 |
| Opus 4.8 | 87.8 |
| Gemini 3.5 Flash | 87.6 |
| DeepSeek V4 Pro 0813 | 87.5 |
| DeepSeek-V4-Pro | 87.5 |
| Grok 4.5 | 87.4 |
| GPT-5.3-Codex | 87.3 |
| DeepSeek-V4-Flash-0731 | 87.3 |
| Kimi K3 | 87.2 |
| Qwen3.7-Max (source unspecified) | 87.1 |
| Kimi K2.6 | 86.8 |
| GPT-5 Mini | 86.6 |
| GPT-5.1 | 86.5 |
| Gemini 3 Pro Preview | 86.4 |
| Nemotron 3 Ultra 550B A55B (NVFP4) | 86.0 |
| Qwen3.6 Plus (2026-04-02) | 86.0 |
| Inkling-Small | 85.9 |
| GPT-5.6 Terra | 85.9 |
| GPT-5 | 85.9 |
| Muse Spark 1.1 | 85.9 |
| Gemini 3 Flash Preview | 85.6 |
| GPT-5.1-Codex | 85.5 |
| Inkling | 85.5 |
| GPT-5.2 | 85.4 |
| Qwen3.5 Plus (2026-02-15) | 85.3 |
| GPT-5.5 | 85.3 |
| Opus 4.7 | 85.1 |
| GPT-5-Codex | 84.7 |
| Opus 4.6 | 84.7 |
| Grok 4.3 | 84.5 |
| Grok 4.20 | 84.3 |