ChainWorld (Single Turn): leaderboard
ChainWorld (Siu et al., Scale AI and UC Berkeley, 2026) chains two to four compatible OSWorld desktop tasks into one long workload (347 chains) and keeps each task's original evaluator, so a computer-use agent must complete several objectives in one Ubuntu session whose state carries over, acting through pyautogui on full-screen screenshots. In this protocol the whole chain is given in one prompt with 75 steps per task that can be spent anywhere. The score is chain completion, the fraction of chains (0 to 1) in which every task's evaluator passes; higher is better.
Metric: Chain completion (fraction of 347 chains with every task's evaluator passing; the whole chain in one prompt, 75 steps per task pooled). Source: arxiv.org. Saturation forecast: Around August 2028. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 0.15 | #48 |
| 2 | Claude Opus 4.7 | 0.11 | #75 |
| 3 | Kimi K2.5 | 0.05 | #252 |
| 4 | Gemini 3.1 Pro (Preview) | 0.03 | #91 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=chainworld-single-turn · How It Works · Data refreshed daily, snapshot 2026-11-12.