Benchmarks / Agentic Coding
LiveBench Agentic Coding
Contamination-limited monthly-refreshed benchmark; category mean over its tasks.
maintained by LiveBench · livebench.ai · 36 observations on record
Versions
1 version
| Version | Tasks |
|---|---|
| Release 2026-06-25 current livebench-agentic-coding-2026-06-25 | — |
Current leaderboard
Release 2026-06-25 · Accuracy · top 25 of 36 systems
| # | System | Accuracy % |
|---|---|---|
| 1 | GPT 5.6 Terramax effort OpenAI | 68.0% |
| 2 | GPT 5.6 Solmax effort OpenAI | 65.6% |
| 3 | Grok 4.5 xAI | 59.8% |
| 4 | GPT 5.6 Solxhigh effort OpenAI | 56.6% |
| 5 | Claude Opus 4.8xhigh effort Anthropic | 56.1% |
| 6 | GPT 5.6 Lunamax effort OpenAI | 53.8% |
| 7 | GPT 5.4 (2026-03-05)xhigh effort OpenAI | 53.8% |
| 8 | GPT 5.6 Terraxhigh effort OpenAI | 53.3% |
| 9 | GPT 5.5 (2026-04-22)xhigh effort OpenAI | 52.1% |
| 10 | GLM 5.2 Z.ai | 51.9% |
| 11 | Claude Sonnet 5xhigh effort Anthropic | 51.1% |
| 12 | Claude Opus 4.7xhigh effort Anthropic | 50.7% |
| 13 | Claude Fable 5xhigh effort Anthropic | 50.7% |
| 14 | GPT 5.2 (2025-12-11)high effort OpenAI | 50.3% |
| 15 | GPT 5.2 Codex OpenAI | 49.4% |
| 16 | Gemini 3.5 Flashhigh effort Google DeepMind | 49.0% |
| 17 | Claude Opus 4.6high effort Anthropic | 49.0% |
| 18 | GPT 5.6 Lunaxhigh effort OpenAI | 48.8% |
| 19 | Kimi K2 6thinking Moonshot AI | 46.9% |
| 20 | Claude Fable 5max effort Anthropic | 46.9% |
| 21 | GPT 5.5 (2026-04-22)high effort OpenAI | 46.8% |
| 22 | GPT 5.4 Nano (2026-03-17)xhigh effort OpenAI | 46.8% |
| 23 | Grok Build 0.1 xAI | 45.8% |
| 24 | Kimi K2 7 Code Moonshot | 45.7% |
| 25 | Gemini 3.1 Pro Previewhigh effort Google DeepMind | 45.4% |
One row per evaluated system — reasoning-effort variants rank separately.