PitchBench - Pitch under Audio Effects: leaderboard
Pitch hearing in audio-language models (PitchBench, Liessens Dujardin et al., 2026): naming the pitch of a tone after filtering, bitcrushing, soft clipping, reverb or chorus (240 stimuli), with synthetic stimuli. A pitch answer counts when any of MIDI number, scientific pitch notation or frequency is right; accuracy in percent, higher is better.
Metric: Accuracy (%; name the pitch after filtering, bitcrushing, saturation, reverb or chorus). Source: arxiv.org. Saturation forecast: Around 2029. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 Omni Plus | 86.7 | #237 |
| 2 | Qwen3.5 Omni Flash | 55.8 | #548 |
| 3 | Gemini 3.1 Pro (Preview) | 21.3 | #91 |
| 4 | Gemini 3 Flash | 19.6 | #159 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=pitchbench-pitch-under-audio-effects · How It Works · Data refreshed daily, snapshot 2026-11-12.