ZONOS3-Mini
A compact 800M mixture-of-experts voice-cloning model with 100M active parameters. Apache 2.0, 43 languages, runs in a browser tab.
In short. ZONOS3-Mini reads text aloud in 43 languages and can copy a voice from a short recording. Zyphra says it is small enough to run in real time on a laptop or phone, even inside a web browser with no server. The weights are Apache 2.0, so you can use them in products, but you still need permission from anyone whose voice you clone.
Summary
ZONOS3-Mini is a text-to-speech model from Zyphra, a San Francisco AI lab, released on 9 November 2026 under the Apache 2.0 license. It speaks 43 languages and clones a voice from a reference clip of up to 60 seconds or from a stored speaker embedding. Zyphra built it to run in real time on phones and laptops, including inside a browser tab through WebGPU.
It is a mixture-of-experts model with 800 million parameters in total and about 100 million active per token. It has 11 layers with a hidden size of 1,024, a latent mixture of 64 experts with 2 chosen per token, and a 3-to-1 mix of Gated DeltaNet and global-attention layers. The language model predicts audio frames at 15 per second in Zyphra's own codec, ZAC-1, which uses ten codebooks (one semantic and nine acoustic) and a causal decoder that renders 44.1 kHz audio as it streams. One generation can last up to 10 minutes. Zyphra says it trained the model on 112 NVIDIA H100 GPUs in 2.5 days, and that post-training used preference optimization followed by GRPO. Besides cloning, it has controls for age, gender, accent and voice texture, steering for pace, emotion and delivery down to single words, and an edit model that rewrites part of a finished sentence.
Zyphra reports a median time to first audio of 80 ms with 8 concurrent requests on one H100 GPU and a peak throughput of 457 seconds of audio per second. In a browser it reports under 570 ms on a Pixel 9 and under 150 ms on an M3 MacBook Air. On its own Zyphra TTS Capability Index, built from its ZTTS2-Eval benchmark of 55 language subsets, ZONOS3-Mini ranks first among open models, ahead of Fish Audio S2 Pro and MOSS TTS v1.5. The benchmark and the index are Zyphra's own. The language-model weights are 1.8 GB in BF16, and the PyTorch weights that the CUDA server needs, including the audio codec, are about 9 GB. Zyphra ships runtimes for NVIDIA GPUs with SGLang-Omni, for Apple Silicon with MLX, for CPU, CUDA, Metal and Vulkan with a C++ ggml server, and for browsers with WebGPU, and it also sells a hosted version on Zyphra Cloud.
Good for
- Voice cloning and narration in many languages on a laptop or phone, including fully in the browser with no server
- Games and assistants that need controllable delivery, such as accent, age, emotion and emphasis on single words
- Streaming speech with a short time to first audio on a server GPU, by Zyphra's own measurement
- Products that need plain Apache 2.0 terms for the weights
Watch out for
- Cloning a real person's voice without permission can be illegal or harmful, and the Apache 2.0 license does not change that
- The quality ranking comes from Zyphra's own benchmark, so listen to your language and voices before you choose
- Text normalization covers only nine languages, and other languages pass through as written, so numbers and abbreviations may be read badly
- The Hugging Face repository is about 31 GB because it holds training files, and each runtime needs only specific files (GGUF or web shards)
- It was released on 9 November 2026, so it has little track record
Performance claims come from the developer and have not been independently verified.
How much memory does it need?
Memory figures for each variant of ZONOS3-Mini. Image, video, speech and music pipelines also load a text encoder and a decoder, so the memory needed can be larger than the main checkpoint.
| Variants | Parameters | Download size | Memory needed | Fits on |
|---|---|---|---|---|
| ZONOS3-Mini BF16 | 800M 100M active | 9 GB | 10 GB | Laptop or phone |
Download size covers every component needed to run the model, such as the generator, text encoder and decoder. Memory needed is a stated or published measured minimum where one exists; otherwise it is the download size plus about 15%. Larger images, longer clips and higher resolutions need more.
The license in plain words
One of the most widely used permissive licenses. You may use, change and sell products built on the model with only light conditions.
See also
- ZONOS3-Mini vs Drama 3: which new open TTS model?
- Best open-source speech models for transcription and text-to-speech
- Best open-source text-to-speech models in 2026
- Open speech models: transcription is open, text-to-speech is mixed
- Tiny text-to-speech models that run on a CPU: four open options compared
- ZONOS3-Mini: open TTS in 43 languages that runs in a browser
Frequently asked questions
- What is ZONOS3-Mini?
- ZONOS3-Mini reads text aloud in 43 languages and can copy a voice from a short recording. Zyphra says it is small enough to run in real time on a laptop or phone, even inside a web browser with no server. The weights are Apache 2.0, so you can use them in products, but you still need permission from anyone whose voice you clone.
- Can I use ZONOS3-Mini commercially?
- Yes. ZONOS3-Mini is released under Apache 2.0, a permissive license. You can use it commercially if you follow the standard conditions, such as keeping the license notice.
- Is ZONOS3-Mini open source?
- ZONOS3-Mini is released under Apache 2.0, a license approved by the Open Source Initiative. Training data and training code may not be public, so it is best described as open weights under an open-source license.
- Can I run ZONOS3-Mini on my own computer?
- Yes. The smallest variant, ZONOS3-Mini, needs about 10 GB of memory at 4-bit precision, which fits a modern laptop or a mid-range graphics card.
- When was ZONOS3-Mini released?
- ZONOS3-Mini was released by Zyphra on November 9, 2026.
Official links
Sources
Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.
- ZONOS3-Mini (Zyphra blog, 9 Nov 2026) zyphra.com
- ZONOS3-Mini model card (Hugging Face) huggingface.co
- ZONOS3-Mini runtimes and README (GitHub) github.com
- ZTTS2-Eval benchmark (Zyphra) zyphra.github.io
Release date and facts last verified November 11, 2026.
Related models
| Model | Released | Parameters | Context | License | Smallest variant needs |
|---|---|---|---|---|---|
| Drama 3 Fish Audio | 35.9B 3B active | – | Fish Audio Community 1.0 (Conditional) | 25 GB Gaming GPU | |
| Pocket TTS Kyutai | 18M – 100M | – | CC BY 4.0 (Permissive) | 0.127 GB Laptop or phone | |
| Qwen3-TTS Alibaba Qwen | 910M – 1.92B | – | Apache 2.0 (Permissive) | 3.68 GB Laptop or phone | |
| VoxCPM2 OpenBMB | 2.29B | – | Apache 2.0 (Permissive) | 8 GB Laptop or phone |