ODU-Bench (Audio-Only) - Transcription: leaderboard
Metric: Transcript accuracy (%; max(0, 1 - pooled CER or WER) of the demand transcript, demand scenes only; 731 synthetic audio-only scenes (570 with a demand, 161 without), user turns as audio, earlier assistant replies as text). Source: arxiv.org. Saturation forecast: Around December 2026. 14 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Seed 2.0 Lite | 93.2 |
| 2 | Gemini 3.1 Pro (Preview) | 91.7 |
| 3 | Qwen3 Omni 30B A3B Instruct | 91 |
| 4 | Qwen 3.5 Omni Plus | 89.3 |
| 5 | Gemini 3.7 Flash | 88.4 |
| 6 | Gemini 3.5 Flash Lite | 74.3 |
| 7 | Qwen2.5-Omni-7B | 69.4 |
| 8 | MiniCPM-o-4.5 | 61.5 |
Interactive version: theaggregate.ai/benchmark?slug=odu-bench-audio-only-transcription · How It Works · Data refreshed daily, snapshot 2026-09-26.