OccuBench (Clean Environment): leaderboard
Metric: Completion rate (%) over all 382 tasks in the clean environment (E0); 382 tool-use task instances from 100 professional scenarios, the environment simulated by Gemini-3-Flash-Preview and each trajectory passed or failed by a rubric verifier; thinking mode on, high reasoning effort where configurable; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 (High) | 79.6 |
| 2 | Gemini 3.1 Pro (Preview) | 72.3 |
| 3 | Claude Opus 4.6 (High) | 71.5 |
| 4 | Qwen 3.5 Plus (Thinking) | 69.9 |
| 5 | DeepSeek V3.2 (Thinking) | 69.6 |
| 6 | Claude Opus 4.5 (Thinking) | 65.2 |
| 7 | Claude Sonnet 4.5 (Thinking) | 64.9 |
| 8 | Claude Sonnet 4.6 (Thinking) | 64.4 |
| 9 | Kimi K2.5 (Thinking) | 64.1 |
| 10 | GLM-5 (Thinking) | 62.6 |
| 11 | Claude Opus 4 (Thinking) | 61.3 |
| 12 | Qwen 3.5 Flash (Thinking) | 59.7 |
| 13 | MiniMax-M2.7 | 53.9 |
| 14 | Claude Sonnet 4 (Thinking) | 53.4 |
Interactive version: theaggregate.ai/benchmark?slug=occubench-clean-environment · How It Works · Data refreshed daily, snapshot 2026-10-07.