LiveBench Agentic Coding Average
Code · 2025-05-30
LiveBench leaderboard aggregate for livebench agentic coding average. LiveBench is refreshed over time to reduce contamination and covers reasoning, coding, mathematics, language, data analysis, instruction following, and agentic coding.
Top models (higher is better)
| Model | Score |
|---|---|
| Opus 4.8 | 56.1 |
| GPT-5.4 | 53.8 |
| GPT-5.5 | 52.1 |
| GLM-5.2 | 51.9 |
| Sonnet 5 | 51.1 |
| Opus 4.7 | 50.7 |
| GPT-5.2 | 50.3 |
| GPT-5.2-Codex | 49.4 |
| Opus 4.6 | 49.0 |
| Gemini 3.5 Flash | 49.0 |
| Kimi K2.6 | 46.9 |
| GPT-5.4 Nano | 46.8 |
| Grok Build 0.1 | 45.8 |
| Kimi K2.7 Code | 45.7 |
| Gemini 3.1 Pro Preview | 45.4 |
| Qwen3.7-Max (source unspecified) | 43.6 |
| Sonnet 4.6 | 42.6 |
| DeepSeek-V4-Pro | 42.6 |
| GPT-5.4 Mini | 41.7 |
| Qwen3.6 Plus (2026-04-02) | 41.4 |
| MiniMax M3 | 40.7 |
| Opus 4.5 | 39.7 |
| Qwen3.6 27B | 39.3 |
| DeepSeek-V4-Flash | 37.6 |
| Grok 4.3 | 18.5 |