FitAQA - Temporal Grounding: leaderboard

Metric: Mean temporal IoU (%; localizing the intervals where a described form error occurs; over FitAQA's fitness action-quality videos (30 bodyweight exercises, 38-error taxonomy); temporal grounding task). Source: arxiv.org. Saturation forecast: Around September 2027. 18 models tracked.

Top models

#ModelScore
1GPT-5.547.5
2Gemini 3.1 Pro (Preview)44
3Qwen 3.5 27B (Non-reasoning)37.6
4GPT-5.436.9
5Qwen 3.5 27B (Thinking)36.6
6Qwen 3.5 35B A3B (Thinking)31.7
7Qwen 3.5 35B A3B (Non-reasoning)30.5
8Qwen 3.5 9B (Thinking)30.4
9Qwen 3.5 9B (Non-reasoning)25.7
10Gemma 4 31B24.9
11Qwen 3 VL 8B10.1

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