ODU-Bench (Audio-Only): leaderboard
Metric: Weighted demand-understanding score (%; 0.15 demand detection + 0.60 key-point coverage + 0.10 temporal localization + 0.10 transcription + 0.05 user profile, on 731 synthetic audio-only scenes (570 with a demand, 161 without), user turns as audio, earlier assistant replies as text; key points judged by Qwen3.6-Flash with thinking). Source: arxiv.org. Saturation forecast: Around December 2026. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Seed 2.0 Lite | 77.3 |
| 2 | Gemini 3.1 Pro (Preview) | 75.4 |
| 3 | Qwen 3.5 Omni Plus | 74.7 |
| 4 | Gemini 3.7 Flash | 72.9 |
| 5 | Gemini 3.5 Flash Lite | 62.2 |
| 6 | Qwen3 Omni 30B A3B Instruct | 61.6 |
| 7 | MiniCPM-o-4.5 | 49.1 |
| 8 | Qwen2.5-Omni-7B | 48 |
Interactive version: theaggregate.ai/benchmark?slug=odu-bench-audio-only · How It Works · Data refreshed daily, snapshot 2026-09-26.