LiveCodeBench (Vals Public) - Hard
Code · 2024-03-12
The hard difficulty slice of Vals AI's LiveCodeBench public-dataset evaluation surface.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 79.7 |
| Claude Opus 5 | 77.4 |
| Gemini 3.1 Pro Preview | 76.6 |
| GPT-5.3-Codex | 75.1 |
| Opus 4.8 | 74.6 |
| Gemini 3.6 Flash | 74.6 |
| GPT-5.2-Codex | 74.6 |
| Kimi K3 | 74.3 |
| Grok 4.5 | 73.4 |
| Qwen3.7-Max | 73.4 |
| Gemini 3.5 Flash | 73.1 |
| GPT-5.6 Terra | 72.9 |
| DeepSeek-V4-Pro | 72.6 |
| GPT-5 Mini | 71.7 |
| Kimi K2.6 | 71.4 |
| GPT-5.1 | 71.1 |
| Gemini 3 Pro Preview | 70.9 |
| GPT-5.2 | 70.1 |
| Gemini 3 Flash Preview | 69.7 |
| GPT-5.5 | 69.7 |
| Muse Spark 1.1 | 69.1 |
| Nemotron 3 Ultra 550B A55B | 69.1 |
| GPT-5.1-Codex | 68.9 |
| GPT-5-Codex | 68.6 |
| Inkling | 68.3 |
| GPT-5.4 | 67.7 |
| GPT-5.6 Sol | 67.4 |
| Qwen3.5 Plus (2026-02-15) | 67.1 |
| Qwen3.6 Plus (2026-04-02) | 67.1 |
| Opus 4.7 | 66.9 |
| Sonnet 5 | 66.9 |
| Opus 4.6 | 66.3 |
| GPT-5 | 66.3 |
| GPT-5.4 Nano | 66.0 |
| Grok 4.20 | 66.0 |
| Grok 4.3 | 66.0 |
| GPT-5.1-Codex-Max | 65.4 |
| gpt-oss-120b | 64.6 |
| Kimi K2.5 | 64.3 |
| Opus 4.5 | 64.0 |