Open ASR · AMI-Cleaned WER ↗English short-form; normalization unversioned
8.31% WERQwen/Qwen3-ASR-1.7B-hf
QwenDiscovered Oct 17, 2026
Qwen/Qwen3-ASR-1.7B-hf
2.04B parameters · 52 languages · Encoder: Custom · Decoder: Qwen3
Measured performance
All leaderboards ↗Benchmark results
Word error rate · lower is better
Open ASR · Earnings22-Cleaned-AA-chunked WER ↗English short-form; normalization unversioned
5.84% WERQwen/Qwen3-ASR-1.7B-hf
Open ASR · Gigaspeech-Cleaned WER ↗English short-form; normalization unversioned
7.22% WERQwen/Qwen3-ASR-1.7B-hf
Open ASR · LS Clean WER ↗English short-form; normalization unversioned
1.26% WERQwen/Qwen3-ASR-1.7B-hf
Open ASR · LS Other WER ↗English short-form; normalization unversioned
2.94% WERQwen/Qwen3-ASR-1.7B-hf
Open ASR · SPGISpeech WER ↗English short-form; normalization unversioned
2.57% WERQwen/Qwen3-ASR-1.7B-hf
Open ASR · URGENT2024 WER ↗English short-form; normalization unversioned
9.93% WERQwen/Qwen3-ASR-1.7B-hf
Open ASR · Voice Arena Monsoon WER ↗English short-form; normalization unversioned
3.5% WERQwen/Qwen3-ASR-1.7B-hf
Open ASR · Voxpopuli-AA-Cleaned WER ↗English short-form; normalization unversioned
2.85% WERQwen/Qwen3-ASR-1.7B-hf
Access & cost
0 quotesPrices across providers
No listed provider price
Model specifications
- Developer
- Qwen
- Modality
- Audio
- Context
- Not reported
- Maximum output
- Not reported
- Weights
- Not reported
- First discovered
- Oct 17, 2026
More from Qwen
View all ↗QwenListed Jul 20
Qwen3-ASR-1.7B
Qwen3-ASR-1.7B is the flagship model of the Qwen3-ASR family: multilingual language identification and speech recognition across 30 languages and 22 Chinese dialects, built on Qwen3-Omni. It reaches…
1.7BAudio inputOpen weightsNo reported score
No reported score
QwenListed Jul 20
Qwen3-ASR-0.6B
Qwen3-ASR-0.6B is the compact model of the Qwen3-ASR family: multilingual language identification and speech recognition across 30 languages and 22 Chinese dialects, built on Qwen3-Omni. It targets an…
0.6BAudio inputOpen weightsNo reported score
No reported score
QwenListed Mar 6
Qwen3-TTS-VoiceDesign
● Qwen3-TTS-VoiceDesign is a voice design variant of Qwen3-TTS by Alibaba's Qwen team. Instead of selecting from preset voices, you describe the voice you want in natural language — and the model generates…
No reported score
No reported score
QwenListed Mar 6
Qwen3-TTS
Qwen3-TTS is an advanced text-to-speech model by Alibaba's Qwen team, delivering stable, expressive, and low-latency speech generation across 10 languages. Key capabilities: - 9 preset voices — Vivian,…
No reported score
No reported score