Skip to content

Benchmark

Open Asr Leaderboard

Scores are percentages. Lower is better.

Models measured
8
Scores recorded
8
Sources
1
IQ weight
0.85

Score distribution

6.7%
Mean
7.1%
Median
0.905%
Std dev
5.8%
10th percentile
7.8%
90th percentile
5.4%
Min
5.4%7.9%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.85 for speech models, measured over 7 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Cohere Transcribe 03 2026· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim5.4%
2Qwen3 ASR 1.7B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim5.8%
3Phi 4 Multimodal Instruct· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim6%
4Qwen3 ASR 0.6B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim6.4%
5Voxtral Mini 3B 2507· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim7.1%
6Whisper Large V3· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim7.4%
7Whisper Large v2· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim7.8%
8Whisper Large· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim7.9%

Data credits: Hugging Face official benchmark leaderboards; licence varies by benchmark