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

#ModelScore
1Seed 2.0 Lite93.2
2Gemini 3.1 Pro (Preview)91.7
3Qwen3 Omni 30B A3B Instruct91
4Qwen 3.5 Omni Plus89.3
5Gemini 3.7 Flash88.4
6Gemini 3.5 Flash Lite74.3
7Qwen2.5-Omni-7B69.4
8MiniCPM-o-4.561.5

Interactive version: theaggregate.ai/benchmark?slug=odu-bench-audio-only-transcription · How It Works · Data refreshed daily, snapshot 2026-09-26.