Benchmark Almanac

AISHELL-4

Speaker-Reasoner Multi-turn w/ SAC 14.7% tops its leaderboard
Not in model cards

Models transcribe Mandarin meetings and identify speakers from conference recordings. NVIDIA reports character error after matching speakers.

Made by
AISHELL and collaborators
Released
April 2021
Kind
Speech recognition and translation
Level
Everyday
Scored
Checked automatically
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
Chinese
In and out
Audio → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
In views
Audio and speech
11
Entries on SOTA2 Research’s leaderboard, 0 models we track

Speaker-Reasoner Multi-turn w/ SAC tops the leaderboard with 14.7%

Best score per entry from SOTA2 Research’s leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

  1. 1 Speaker-Reasoner Multi-turn w/ SAC Stage: 3 14.7%
  2. 2 Speaker-Reasoner Multi-turn Stage: 2 14.9%
  3. 3 Speaker-Reasoner Base Stage: 1 16.5%
  4. 4 DM-ASR (S2SND) (CN+EN 2900h) Parms: 1.7B, Training Data / Hours: CN+EN 2900h, Diarization Source: S2SND, Time Prediction Source: LLM-predicted, Label Perturbation Status: None 17.7%
  5. 5 DM-ASR (DiariZen) (CN+EN 2900h) Parms: 0.6B, Training Data / Hours: CN+EN 2900h, Diarization Source: DiariZen, Time Prediction Source: LLM-predicted, Label Perturbation Status: None 19%
  6. 6 DM-ASR (S2SND) (CN 1300h) Parms: 1.7B, Training Data / Hours: CN 1300h, Diarization Source: S2SND, Time Prediction Source: LLM-predicted, Label Perturbation Status: None 19.2%
  7. 7 Qwen3-Omni + SOT sft Architecture: End-to-End, Stage: Ablation baseline 19.6%
  8. 8 DM-ASR (DiariZen) (CN 1300h) Parms: 0.6B, Training Data / Hours: CN 1300h, Diarization Source: DiariZen, Time Prediction Source: LLM-predicted, Label Perturbation Status: None 19.7%
  9. 9 DM-ASR (DiariZen) (CN 630h) Parms: 0.6B, Training Data / Hours: CN 630h, Diarization Source: DiariZen, Time Prediction Source: LLM-predicted, Label Perturbation Status: None 21.6%
  10. 10 Speaker-Reasoner Multi-turn 7B Scale: 7B 22.9%
  11. 11 Speaker-Reasoner Base 7B Scale: 7B 25.6%

Source: SOTA2 Research’s leaderboard

#EntryScoreSettingDateSource
1 Speaker-Reasoner Multi-turn w/ SAC 14.7% Stage: 3 Official leaderboard
2 Speaker-Reasoner Multi-turn 14.9% Stage: 2 Official leaderboard
3 Speaker-Reasoner Base 16.5% Stage: 1 Official leaderboard
4 DM-ASR (S2SND) (CN+EN 2900h) 17.7% Parms: 1.7B, Training Data / Hours: CN+EN 2900h, Diarization Source: S2SND, Time Prediction Source: LLM-predicted, Label Perturbation Status: None Official leaderboard
5 DM-ASR (DiariZen) (CN+EN 2900h) 19% Parms: 0.6B, Training Data / Hours: CN+EN 2900h, Diarization Source: DiariZen, Time Prediction Source: LLM-predicted, Label Perturbation Status: None Official leaderboard
6 DM-ASR (S2SND) (CN 1300h) 19.2% Parms: 1.7B, Training Data / Hours: CN 1300h, Diarization Source: S2SND, Time Prediction Source: LLM-predicted, Label Perturbation Status: None Official leaderboard
7 Qwen3-Omni + SOT sft 19.6% Architecture: End-to-End, Stage: Ablation baseline Official leaderboard
8 DM-ASR (DiariZen) (CN 1300h) 19.7% Parms: 0.6B, Training Data / Hours: CN 1300h, Diarization Source: DiariZen, Time Prediction Source: LLM-predicted, Label Perturbation Status: None Official leaderboard
9 DM-ASR (DiariZen) (CN 630h) 21.6% Parms: 0.6B, Training Data / Hours: CN 630h, Diarization Source: DiariZen, Time Prediction Source: LLM-predicted, Label Perturbation Status: None Official leaderboard
10 Speaker-Reasoner Multi-turn 7B 22.9% Scale: 7B Official leaderboard
11 Speaker-Reasoner Base 7B 25.6% Scale: 7B Official leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Audio and video

All 695 in Audio and video
  • Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy. Qwen 3.8 Max 88.7%
  • LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization. Gemini 4 Argon 91.7%
  • MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy. Qwen 3.8 Max 82.4%
  • MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy. Qwen 3.8 Max 90.8%
  • FLEURS In use Models transcribe read speech in 102 languages built from FLoRes-101 sentences. Model cards usually report word error rate, so lower scores are better. Phi-4-multimodal 4.00%
  • VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules. Qwen3.5-Omni-Plus 93.1%

Data audit

All 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the official leaderboard (11 entries).
  • Pass Models scored 11 entries on its leaderboard, none a model we track.
  • Note Newest score No scores yet.
  • Note Frontier models No models scored yet.
  • Note Headroom Lower is better, so there’s no ceiling to reach.
Leaderboard import 3 pass
  • Pass Importer We read the board on SOTA2 Research: 11 entries at the last read, none a model we track.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check