whatmodel.sh / tts / ZONOS2 8B
ZONOS2 8B
↗ share
+ contribute
ZONOS2 8B is an open-weights text-to-speech model from Zyphra, released June 2026. Its published size is 8B parameters (0.9B active per token). Licence: apache-2.0.
Catalogue identity notes
Zyphra/ZONOS2 (8B total, 900M active MoE). Default mode results; Quality Mode values kept as metadata. Distinct from Zonos v0.1.
Model information sources
Score within the Vapi Humanness · Bradley–Terry rating, 2026-09-25 snapshot comparison set only.
$ how it compares
The text-to-speech models ranked either side of ZONOS2 8B, out of 148 with a score, then more from Zyphra.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 61 |  TTS-1 HD | 57±7 | api $30.00/M chars | ||
| 62 |  Deepgram Aura 2 | 57±3 | api $30.00/M chars | ||
| 63 | ZONOS2 8B | 57±2 | local | ||
| 64 |  Supertonic 3 | 55±10 | local | ||
| 65 |  ElevenLabs v3 Alpha | 55±9 | comparison only | ||
| more from Zyphra | |||||
| 114 | Zonos v0.1 | 39±4 | api price unknown +local | ||
see all text-to-speech models · find the ones that fit your machine
$ what it's good at
voice cloningnot ranked on this task yet · 2 tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Seed-TTS test-en · Whisper large-v3 WER lab | 2.1% ✎ | ▁···▂▂▂▂█▆▄▂ beats 12/35 | — |
| Seed-TTS test-zh · Paraformer CER lab | 2.5% ✎ | ▁·▂▃▁▁▂▃▅█·▂ beats 3/30 | — |
other tests5 not in a task · 98% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| ZTTS1-Eval in-the-wild · 10-language mean WER lab | 6.1% ✎ | ▄·····█··▄██ beats 0/7 | 26% |
| ZTTS1-Eval Clean (FLEURS-R) · 9-language mean WER lab | 9.3% ✎ | ▄·····▄█▄█·▄ beats 0/7 | 22% |
| ZTTS1-Eval in-the-wild · 10-language mean MSR-UTMOS lab | 2.40 / 5 ✎ | █··▄█▄····▄▄ beats 1/7 | 20% |
| ZTTS1-Eval Clean (FLEURS-R) · 9-language mean MSR-UTMOS lab | 3.13 / 5 ✎ | ▃▃█···▃···▃▃ beats 1/7 | 19% |
| ZTTS1-Eval · FLEURS-R English TTSDS2 prosody lab | 87.5% ✎ | █··███·····█ beats 1/4 | 10% |
$ where to run it
Other ways to run it
version
- Run it as published · needs ~5.17 GB · from github.com/Zyphra (official)
git clone --recurse-submodules https://github.com/Zyphra/zonos2.cpp.git && cd zonos2.cpp && cmake -B build -DCMAKE_BUILD_TYPE=Release && cmake --build build -j && hf download Zyphra/ZONOS2-GGUF dac.gguf spk-encoder.gguf --local-dir out && hf download Zyphra/ZONOS2-GGUF zonos2-q4_k.gguf --local-dir out && ./build/zonos2-cli out/zonos2-q4_k.gguf --tts "Hello, world." out.wav --dac out/dac.gguf --seed 1
Official weights and documentation: huggingface.co/Zyphra ↗
$ how the score was worked out · ZONOS2 Technical Report (ZTTS… 87%, ZTTS1-Eval results snapshot (… 10%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 ZONOS2 Technical Report (ZTTS1-Eval results) · lab-reported | 4 | 87% |
 ZTTS1-Eval results snapshot (Zyphra) · lab-reported | 1 | 10% |
| prior · pull toward the middle | 2% | |
| weighted average | = 56.5 |
not counted (2)
| ZONOS2 Technical Report (ZTTS1-Eval results) | Seed-TTS test-en · Whisper large-v3 WER [correct] | 2.1% | fewer than three models have this test yet, or this source ran it for this model alone |
| ZONOS2 Technical Report (ZTTS1-Eval results) | Seed-TTS test-zh · Paraformer CER [correct] | 2.5% | fewer than three models have this test yet, or this source ran it for this model alone |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json