OccuBench - Science & Research: leaderboard

Metric: Completion rate (%) on the Science & Research industry's 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

#ModelScore
1GPT-5.2 (High)94
2Gemini 3.1 Pro (Preview)81
3Claude Opus 4 (Thinking)81
4Claude Sonnet 4.6 (Thinking)69
5Qwen 3.5 Plus (Thinking)69
6Qwen 3.5 Flash (Thinking)69
7MiniMax-M2.762
8DeepSeek V3.2 (Thinking)62
9Claude Opus 4.6 (High)62
10GLM-5 (Thinking)62
11Kimi K2.5 (Thinking)56
12Claude Opus 4.5 (Thinking)56
13Claude Sonnet 4.5 (Thinking)44
14Claude Sonnet 4 (Thinking)31

Interactive version: theaggregate.ai/benchmark?slug=occubench-science-research · How It Works · Data refreshed daily, snapshot 2026-10-07.