HolyGrAIL

All tasks / Models

Qwen3-TTS-12Hz-1.7B

Alibaba's Apache-2.0 open-weights text-to-speech models (0.6B and 1.7B), with custom-voice and voice-design variants.

  • Alibaba (Qwen)
  • Open weights
  • Released 21 Jan 2026

Official page

Where it ranks

  • –
    Text to speech
    Models · Local · no result on this task’s benchmarks yet

Facts

Vendor
Alibaba (Qwen)
Family
Qwen3-TTS
Released
21 Jan 2026
Weights
Open: download and run it yourself
License
Apache-2.0
Parameters
1.93B
GPU memory to run locally
~6 GB (Derived: 1.93B parameters (Hugging Face safetensors count, 1.7B-Base) x 2 bytes = ~3.9 GB of bf16 weights, plus overhead.)
Access
Free (open weights)
Last verified
18 Oct 2026

Benchmark results

No public benchmark results recorded yet.

Where you can use it

Sources