Zyphra

weights

open(ZONOS2 (model.pth) and GGUF conversions on the Hub, ungated)

data

described(more than six million hours of web-scale multilingual speech

code

partial(inference server and offline API

license

Apache-2.0(ZONOS2 weights per the Hub card and Zyphra's announcement

superseded-release

Zonos-v0.1(2025 transformer and hybrid models, Apache-2.0, still published)

ZONOS2, the current release, downloads freely under Apache-2.0, as did Zonos v0.1 before it. Zyphra describes its six million hours of training audio without naming or releasing them and publishes only inference code.

• https://huggingface.co/api/models?author=Zyphra&limit=100 recorded 2026-09-27

  Hub listing: ZONOS2 created 2026-06-11 and Zonos-v0.1-transformer and -hybrid created 2025-02-06, all tagged "license:apache-2.0".

• https://huggingface.co/api/models/Zyphra/ZONOS2 recorded 2026-09-27

  Hub record: tag "license:apache-2.0", "gated":false, and model.pth in the file list.

• https://raw.githubusercontent.com/Zyphra/ZONOS2/HEAD/LICENSE recorded 2026-09-27

  Code LICENSE: "MIT License" / "Copyright (c) 2026 Zyphra, Inc."

• https://raw.githubusercontent.com/Zyphra/ZONOS2/HEAD/README.md recorded 2026-09-27

  README: "ZONOS2 is our latest text-to-speech model trained on more than 6 million hours of varied multilingual speech"; the repository documents a server and offline inference only.

• https://www.zyphra.com/our-work/zonos2 recorded 2026-09-27

  Announcement dated "Jun 12, 2026": "Zyphra has released ZONOS2, an Apache 2.0-licensed real-time text-to-speech model"; "The ZONOS2 model weights are available on Hugging Face"; "we increased our training set from roughly 200,000 hours to over six million hours".

Hugging Face downloads summed over the declared checkpoints, still mostly the 2025 v0.1 transformer model.

• https://huggingface.co/api/models?author=Zyphra&limit=100 recorded 2026-09-27

  Trailing-30-day downloads: Zonos-v0.1-transformer 107,508, ZONOS2-GGUF 10,615, Zonos-v0.1-hybrid 1,196, ZONOS2 692.

The new ZONOS2 has no listener rating yet, so it is placed on its own results on the shared Seed-TTS test set, level with Kokoro. Its predecessor sits in the lower part of the arena, and a rating for ZONOS2 would replace this reading.

• https://artificialanalysis.ai/text-to-speech/leaderboard recorded 2026-09-27

  Leaderboard entry "name":"Zonos-v0.1" (creator Zyphra) with "openWeights":true.

• https://www.zyphra.com/our-work/zonos2 recorded 2026-09-27

  Announcement: "competitive with leading open-source TTS models on widely used open benchmarks such as seed-tts eval".