FMSU-Bench - Pitch (Chinese): leaderboard

FMSU-Bench (2026) tests fine-grained speech understanding in Chinese and English with multiple-choice questions whose distractors include subtle acoustic confusions and text-only semantic traps, plus a tagged transcription task. This board is accuracy on identifying the pitch characteristics in the Chinese subset; percent, higher is better.

Metric: Accuracy (%) of identifying the pitch characteristics (PIT), on the FMSU-Bench Chinese subset (500 to 1,000 manually verified items per task, from audiovisual sources), multiple-choice questions whose distractors include fine-grained acoustic confusions and text-only semantic traps; higher is better. Source: arxiv.org. Saturation forecast: Around January 2028. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash69.3#175
2Qwen 3 Omni 30B A3B68.2#847
3Gemini 3.1 Pro (Preview)66.6#99
4Qwen2.5-Omni-7B63.8#1884
5Gemini 2.5 Flash62.1#463
6MiMo-Audio59.8#931
7Kimi-Audio37.2#1735
8FMSU-Bench Step-Audio 2 (checkpoint unspecified)24.3
9Audio Flamingo 320.2
10Qwen2-Audio (FMSU-Bench checkpoint unspecified)15.8

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=fmsu-bench-pitch-chinese · How It Works · Data refreshed daily, snapshot 2026-11-21.