WatchAct - Imitation: 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; 290 instances). Source: arxiv.org. Saturation forecast: Around September 2027. 9 models tracked.

Top models

#ModelScore
1GPT-5.133.4
2GPT-5.432.8
3Gemini 3.1 Pro (Preview)30
4Gemini 3 Flash24.5
5Qwen 3 VL 235B A22B Instruct12.8
6Qwen 3 VL 235B A22B (Thinking)9.7
7Gemini 3.1 Flash Lite9

Interactive version: theaggregate.ai/benchmark?slug=watchact-imitation · How It Works · Data refreshed daily, snapshot 2026-09-26.