WatchAct - Temporal Sort: 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; 210 instances). Source: arxiv.org. Saturation forecast: Around September 2027. 9 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)28.1
2GPT-5.427.6
3GPT-5.126.5
4Gemini 3 Flash22.2
5Qwen 3 VL 235B A22B Instruct8.6
6Qwen 3 VL 235B A22B (Thinking)7.2
7Gemini 3.1 Flash Lite6.1

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