Model · Alibaba · Speech-To-Text
Qwen3-ASR
Alibaba's open speech-recognition family. The 1.7B model transcribes 30 languages and 22 Chinese dialects and detects the language itself; a separate ForcedAligner-0.6B adds word-level timestamps.
- Modality
- Speech-To-Text
- License
- Apache 2.0 (Open)
- Parameter size
- 1.7B (headline); 0.6B; ForcedAligner 0.6B
- Released
- January 29, 2026
- Last verified
- October 28, 2026
- Runs locally
- Yes
What you need to run it
- Mac
- Yes. Any Apple Silicon Mac (M1 or newer) with at least 8 GB (16 GB recommended) of memory. Newer chips are faster, but memory is what decides whether it runs.
- Windows / Linux PC
- NVIDIA works best. A graphics card with at least 6 GB (12 GB recommended) of video memory.
- No graphics card
- Not practical. Too slow without a graphics card or Apple Silicon.
Mac users run it through MLX builds (mlx-community, mlx-audio); the official code assumes a CUDA graphics card.
Numbers are for: 0.6B for the minimum, 1.7B (bf16) for the recommended. Our estimate from the model’s size — the maker publishes no figure.
Strengths
- Apache 2.0 — no commercial restrictions
- 30 languages plus 22 Chinese dialects, with automatic language identification
- Also recognises singing and speech over music
- Timestamps for any unit (word, character) via Qwen3-ForcedAligner-0.6B, up to 5 minutes per clip in 11 languages
- vLLM backend for batch, async and streaming serving; the 0.6B model reaches 2000x throughput at 128 concurrent requests
- Native Hugging Face Transformers support since 2026-06-26
Weaknesses
- Timestamps need the separate aligner model, which covers 11 languages, not all 30
- No speaker diarization — it does not say who spoke
Try it
| Where | Type | Notes |
|---|---|---|
| Hugging Face | weights | Apache 2.0; 1.7B, 0.6B and ForcedAligner-0.6B |
| GitHub (qwen-asr package) | local | pip install qwen-asr; optional vLLM backend; Docker image |
Official sources
- Model card (1.7B) model
- GitHub github
- Qwen blog announcement
Change log
- — Added hardware requirements (Mac support, memory, NVIDIA). Source: GitHub README + mlx-community/Qwen3-ASR-1.7B-4bit + size.
- — Initial entry. Added 2026-10-28 from the popularity sweep: 1.39M Hugging Face downloads/month for the 1.7B model; 3.8k GitHub stars. Released 2026-01-29 per the GitHub README news.
Esc