All tasks / Models
Qwen3-ASR-1.7B
Alibaba's Apache-2.0 open-weights speech recognition models from the Qwen team (0.6B and 1.7B sizes).
- Alibaba (Qwen)
- Open weights
- Released Jan 2026
Where it ranks
- ATranscriptionModels · Local · 5.4% on Hugging Face Open ASR Leaderboard (English, short-form average WER) (secondary benchmark)
Facts
- Vendor
- Alibaba (Qwen)
- Family
- Qwen3-ASR
- Released
- Jan 2026
- Weights
- Open: download and run it yourself
- License
- Apache-2.0
- Parameters
- 2.35B
- GPU memory to run locally
- ~6 GB (Derived: 2.35B parameters (Hugging Face safetensors count, including the audio encoder) x 2 bytes = ~4.7 GB of bf16 weights.)
- Access
- Free (open weights)
- Last verified
- 18 Oct 2026
Benchmark results
| Benchmark | Result | Entry on the leaderboard | As of | Source |
|---|---|---|---|---|
| Coval Benchmarks: Speech-to-Text WER (24h window) | 3.2% | baseten / qwen3-asr-1.7bCoval STT benchmark, mean WER over the trailing 24h window (snapshot generation 2610, as_of 2026-10-19T09:38Z), n=456 clips; rank 7 of 31; read from the public API that feeds https://benchmarks.coval.ai. | 19 Oct 2026 | benchmarks-api-6wxgp27p2a-ue.a.run.app |
| Hugging Face Open ASR Leaderboard (English, short-form average WER) | 5.4% | Qwen/Qwen3-ASR-1.7B-hfAverage WER over the board's default datasets (English short-form, version 17-10-2026); rank 9; read from the Space's initial table data. | 17 Oct 2026 | huggingface.co |
Where you can use it
- –MuesliMuesli (open source)