AudioProcessBench: leaderboard
Metric: Overall score (0-100): mean of the step-correctness, error-type and chain-aggregation averages for an audio model prompted as a step-level critic; 3,872 reasoning traces (23,497 steps, 41% erroneous) from six audio and omni models on MMAR, MMSU and MMAU-Pro questions, steps labelled by two LLM annotators with human review; few-shot critic prompt with three annotated example chains, temperature 0.7; random 37.2; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash | 67.9 |
| 2 | Gemma 4 E4B | 50.5 |
| 3 | Gemma 3n E4B | 49.7 |
| 4 | Gemma 3n E2B | 45.6 |
| 5 | Gemma 4 E2B | 43.6 |
| 6 | Phi-4 Multimodal Instruct | 43.2 |
| 7 | Qwen2.5-Omni-7B | 42.1 |
Interactive version: theaggregate.ai/benchmark?slug=audioprocessbench · How It Works · Data refreshed daily, snapshot 2026-09-29.