PitchBench - Sequential Interval: leaderboard

Pitch hearing in audio-language models (PitchBench, Liessens Dujardin et al., 2026): giving the signed semitone interval between two successive notes (236 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 (%; give the signed interval between two successive notes). Source: arxiv.org. Saturation forecast: Around November 2027. 6 models tracked.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)19.5#91
2Qwen 3.5 Omni Plus15.7#237
3Qwen3.5 Omni Flash5.5#548
4Gemini 3 Flash5.1#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-sequential-interval · How It Works · Data refreshed daily, snapshot 2026-11-12.