Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Zyphra Profile picture
Zyphra
@ZyphraAI
Open superintelligence
May 6 • 11 tweets • 5 min read
Today we're releasing ZAYA1-8B, a reasoning MoE trained on @AMD and optimized for intelligence density.

With <1B active params, it outperforms open-weight models many times its size on math and reasoning, closing in on DeepSeek-V3.2 and GPT-5-High with test-time compute. 🧵 Image On math and code, ZAYA1-8B beats every model in its SLM weight class, is ahead of Qwen3.5-4B and Gemma-4-E4B and competitive with first-gen frontier reasoning models like DeepSeek-R1-0528, Gemini-2.5-Pro, and Claude 4.5 Sonnet on challenging mathematical reasoning tasks. Image
Apr 10, 2025 • 5 tweets • 3 min read
Zyphra is releasing our first reasoning model, ZR1-1.5B. This small but powerful reasoning model excels at both math and code, making it one of the best models in these categories for its size. It also uses 60% less reasoning tokens than comparable models.

🆓Apache 2.0 license. Image On hard coding evaluations, ZR1-1.5B achieves rough parity with Claude3-Opus and Gemma2-27B-instruct and improves over the base R1-Distill-1.5B model by over 50%. ZR1-1.5B is SoTA on coding evaluations for its size, greatly outperforming code reasoning models such as OpenHands. Image
Feb 10, 2025 • 4 tweets • 2 min read
Today, we're excited to announce a beta release of Zonos, a highly expressive TTS model with high fidelity voice cloning.

We release both transformer and SSM-hybrid models under an Apache 2.0 license.

Zonos performs well vs leading TTS providers in quality and expressiveness. Try the model in our playground and build with our model API: playground.zyphra.com/audio

Read our blog post: zyphra.com/post/beta-rele…

Get the weights on Huggingface: huggingface.co/Zyphra/Zonos-v… and huggingface.co/Zyphra/Zonos-v…

Download the inference code: github.com/Zyphra/Zonos
Oct 15, 2024 • 8 tweets • 3 min read
Another release! With @NvidiaAI we made the best open-source pre-training dataset: Zyda-2.

Zyda-2 is a 5T token dataset distilled from high-quality sources like Zyda-1, DCLM, FineWeb-Edu, and more. Deduplicated/filtered with GPUs thanks to NVIDIA NeMo Curator.

👇Read more👇 Image We built Zyda-2 with the following pipeline:

1) Cross-deduplication of DCLM, Zyda-1, and Dolma-CC v1.7

2) Model-based filtering of Zyda-1 and Dolma-CC

3) Filter FineWeb-Edu2 by its educational score Image