OccuBench (Explicit Faults): leaderboard
Metric: Completion rate (%) over all 382 tasks when the simulator injects explicit faults (HTTP 500, timeouts; E1, two transient fault events of two calls each); 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. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 (High) | 75.9 |
| 2 | Gemini 3.1 Pro (Preview) | 73.3 |
| 3 | Claude Opus 4.6 (High) | 68.1 |
| 4 | Claude Sonnet 4.6 (Thinking) | 62.8 |
| 5 | Qwen 3.5 Plus (Thinking) | 61 |
| 6 | DeepSeek V3.2 (Thinking) | 59.9 |
| 7 | GLM-5 (Thinking) | 59.4 |
| 8 | MiniMax-M2.7 | 52.9 |
| 9 | Kimi K2.5 (Thinking) | 50 |
Interactive version: theaggregate.ai/benchmark?slug=occubench-explicit-faults · How It Works · Data refreshed daily, snapshot 2026-10-07.