Skip to content

Benchmark

OmniCapBench · Structural · Subject reference F1

Scores are percentages. Higher is better.

Models and systems measured
9
Scores recorded
9
Sources
1
IQ weight
No weight published

Score distribution

62.8%
Mean
65.3%
Median
16.3%
Std dev
54.4%
10th percentile
78.6%
90th percentile
24.4%
Min
24.4%80.7%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#Model or systemScore
1Gemini 3.1 Pro PreviewOmniCapBench benchmark-owner run · 786 audio-video clips (12.8 hours) · structured MTSS captions · macro average across videos · rule-based structural track with LLM-assisted reference matching · candidate reasoning effort, exact API snapshot and judge model not specified on public leaderboard· Reported by OmniCapBench authors / Tencent Hunyuan· via BenchBoard docs78.6%
2Qwen3.5 Omni PlusOmniCapBench benchmark-owner run · 786 audio-video clips (12.8 hours) · structured MTSS captions · macro average across videos · rule-based structural track with LLM-assisted reference matching · candidate reasoning effort, exact API snapshot and judge model not specified on public leaderboard· Reported by OmniCapBench authors / Tencent Hunyuan· via BenchBoard docs73.6%
3Qwen3.5 Omni FlashOmniCapBench benchmark-owner run · 786 audio-video clips (12.8 hours) · structured MTSS captions · macro average across videos · rule-based structural track with LLM-assisted reference matching · candidate reasoning effort, exact API snapshot and judge model not specified on public leaderboard· Reported by OmniCapBench authors / Tencent Hunyuan· via BenchBoard docs72.3%
4MiMo V2.5OmniCapBench benchmark-owner run · 786 audio-video clips (12.8 hours) · structured MTSS captions · macro average across videos · rule-based structural track with LLM-assisted reference matching · candidate reasoning effort, exact API snapshot and judge model not specified on public leaderboard· Reported by OmniCapBench authors / Tencent Hunyuan· via BenchBoard docs61.2%

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
Gemini 2.5 ProOmniCapBench benchmark-owner run · 786 audio-video clips (12.8 hours) · structured MTSS captions · macro average across videos · rule-based structural track with LLM-assisted reference matching · candidate reasoning effort, exact API snapshot and judge model not specified on public leaderboard· Reported by OmniCapBench authors / Tencent Hunyuan· via BenchBoard docs80.7%
MiniCPM-o 2.6 (OmniCapBench label)OmniCapBench benchmark-owner run · 786 audio-video clips (12.8 hours) · structured MTSS captions · macro average across videos · rule-based structural track with LLM-assisted reference matching · candidate reasoning effort, exact API snapshot and judge model not specified on public leaderboard· Reported by OmniCapBench authors / Tencent Hunyuan· via BenchBoard docs24.4%
Qwen3 Omni Captioner (OmniCapBench label)OmniCapBench benchmark-owner run · 786 audio-video clips (12.8 hours) · structured MTSS captions · macro average across videos · rule-based structural track with LLM-assisted reference matching · candidate reasoning effort, exact API snapshot and judge model not specified on public leaderboard· Reported by OmniCapBench authors / Tencent Hunyuan· via BenchBoard docs54.4%
Qwen3 Omni Instruct (OmniCapBench label)OmniCapBench benchmark-owner run · 786 audio-video clips (12.8 hours) · structured MTSS captions · macro average across videos · rule-based structural track with LLM-assisted reference matching · candidate reasoning effort, exact API snapshot and judge model not specified on public leaderboard· Reported by OmniCapBench authors / Tencent Hunyuan· via BenchBoard docs54.8%
Seed2.0 (OmniCapBench label)OmniCapBench benchmark-owner run · 786 audio-video clips (12.8 hours) · structured MTSS captions · macro average across videos · rule-based structural track with LLM-assisted reference matching · candidate reasoning effort, exact API snapshot and judge model not specified on public leaderboard· Reported by OmniCapBench authors / Tencent Hunyuan· via BenchBoard docs65.3%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.