FMSU-Bench - Tone (English): 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 tonal characteristics of the voice in the English subset; percent, higher is better.

Metric: Accuracy (%) of identifying the tonal characteristics of the voice (TON), on the FMSU-Bench English 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 August 2027. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Flash58.2#463
2Qwen 3 Omni 30B A3B57.6#847
3Gemini 3 Flash56.6#175
4Gemini 3.1 Pro (Preview)55.6#99
5MiMo-Audio54.6#931
6Kimi-Audio42.6#1735
7FMSU-Bench Step-Audio 2 (checkpoint unspecified)41.7
8Qwen2.5-Omni-7B40.7#1884
9Audio Flamingo 336.2
10Qwen2-Audio (FMSU-Bench checkpoint unspecified)13.9

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-tone-english · How It Works · Data refreshed daily, snapshot 2026-11-21.