EntCollabBench - Workflow: leaderboard
Metric: Task success (%) on the 200 operational workflow tasks (160 single-step and 40 multi-step), judged from execution traces and service-state diffs, role-specialized LLM agents (11 enterprise roles behind permission-controlled ITSM, HR, CSM, Gitea and document services) delegate over HTTP, temperature 0, three-judge majority vote; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Pro | 59 |
| 2 | DeepSeek V4 Flash | 51 |
| 3 | Claude Sonnet 4.6 | 44 |
| 4 | Gemini 3.1 Pro (Preview) | 42.5 |
| 5 | GPT-5.4 | 33.5 |
| 6 | Qwen 3.5 122B A10B | 27.5 |
| 7 | Qwen 3.5 35B A3B | 25 |
| 8 | MiMo-V2-Flash | 21.5 |
| 9 | MiniMax-M2.7 | 18.5 |
| 10 | GPT-5 Mini | 17 |
| 11 | Gemini 3.1 Flash Lite (Preview) | 2 |
| 12 | Qwen 3.5 9B | 0 |
Interactive version: theaggregate.ai/benchmark?slug=entcollabbench-workflow · How It Works · Data refreshed daily, snapshot 2026-10-07.