WatchAct - State Change: leaderboard
Metric: Plan success rate (%; the VLM's high-level Pick/Place/Open/Close plan, predicted from a real human video and an instruction, run through a symbolic executor and checked against the task's success predicate; 155 instances). Source: arxiv.org. Saturation forecast: Around December 2026. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 | 60.6 |
| 2 | Gemini 3.1 Pro (Preview) | 60 |
| 3 | GPT-5.1 | 54.6 |
| 4 | Gemini 3 Flash | 53.5 |
| 5 | Qwen 3 VL 235B A22B (Thinking) | 31 |
| 6 | Qwen 3 VL 235B A22B Instruct | 14.8 |
| 7 | Gemini 3.1 Flash Lite | 14.2 |
Interactive version: theaggregate.ai/benchmark?slug=watchact-state-change · How It Works · Data refreshed daily, snapshot 2026-09-26.