TraceAV-Bench: leaderboard

TraceAV-Bench (2026) tests multi-hop trajectory reasoning over long audio-visual videos: four-option questions whose answers chain several timestamped audio and visual evidence hops across videos of 10 to 140 minutes. The score is accuracy in percent, the macro-average of 12 general sub-tasks with full audio and video input; higher is better.

Metric: Accuracy (%), macro-average of the 12 general sub-tasks, four-option multiple choice over long audio-visual videos (10 to 140 minutes), exact match of the selected option set, each model at its official inference setting, with the full audio and video input; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 15 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)68.29#102
2Gemini 2.5 Pro63.52#280
3Gemini 3 Flash62.28#180
4Gemini 2.5 Flash59.04#466
5Ming-Flash-Omni-2.051.7#1258
6Gemini 2.0 Flash51.59#731
7Qwen 3 Omni 30B A3B48.43#851
8OmniVinci-9B47.38
9MiniCPM-o-4.544.83#1251
10Qwen2.5-Omni-7B42.06#1888
11Gemma 4 E4B40.93#2199
12video-SALMONN 238.63#2695
13HumanOmni-7B38.5#3241
14VideoLLaMA2.1-7B-AV30.43#4199
15Baichuan-Omni-1.526.91#3189

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=traceav-bench · How It Works · Data refreshed daily, snapshot 2026-11-22.