Full stack open superintelligence
San Francisco, CA
Joined March 2021
- Tweets399
- Following40
- Followers15K
- Likes599
Meet ZONOS3-Mini: open-source text-to-speech with voice cloning and word-level control. All on-device and super fast.
Every voice in this video is from a TTS model that can run real-time on a phone.
Comment “ZONOS” within 2 days for 20 hours of free audio.
Sound on 🔊
🤖 Made with AI
Give your next character a voice.
Zyphra Inference: cloud.zyphra.com/models/zyph…
Browser Playground: zyphra.com/zonos3mini/localp…
Blog: zyphra.com/our-work/zonos3-m…
Weights: huggingface.co/Zyphra/ZONOS3…
Code: github.com/zyphra/zonos3-min…
ZTTS Evaluation: zyphra.github.io/ZTTS2-Eval/
@ZyphraAI is a full stack superintelligence company based in San Francisco on a mission to enable sovereign AI through open models and multisilicon infrastructure.
Apply to join us! zyphra.com/careers
3
789
Our mission at Zyphra is to make open superintelligence accessible to everyone.
Today, we're excited to beta release MAIA, the application layer of our fullstack.
Meet your new colleague. A proactive, multiplayer AI agent for teams.
Free during beta: zyphra.com/maia
Your team's next project should start with what it already knows.
Ask MAIA why a decision was made months ago. Get the reasoning, the people involved, and the sources. Turn that context into a plan, a shared project board, or a recurring progress brief.
MAIA takes action across your tools. Ask it to set up a client kickoff: book the time, draft the invite, and share the notes afterward.
It's proactive, too. A morning briefing. Context before your next meeting. A heads-up when a new discussion conflicts with the plan, while there's still time to act.
Your whole team works with the same MAIA, with shared memory, context, and dynamic apps you can open, revise, and build on together. Start on desktop or in Slack. The same team context carries across both.
You stay in control. MAIA follows your existing permissions, asks before irreversible actions, and lets you view or delete what it remembers. You choose which conversations and apps to share.
1
6
1,091
More on why we're building MAIA: zyphra.com/our-work/introduc…
@ZyphraAI is a full stack superintelligence company based in San Francisco on a mission to enable sovereign AI through open models and multisilicon infrastructure.
Apply to join us! jobs.ashbyhq.com/zyphra
1
4
578
ZYFA is part of the attention engine of the Zyphra Optimization Engine (ZOE) on AMD MI355X, which powers Zyphra Cloud and can be deployed on-prem.
Paper: arxiv.org/abs/2611.07496
Blog: zyphra.com/our-work/zyfa
Code: github.com/Zyphra/zyfa
ZOE: zyphra.com/optimization-engi…
@ZyphraAI is a full stack superintelligence company based in San Francisco on a mission to enable sovereign AI through open models and multisilicon infrastructure.
Apply to join us! zyphra.com/careers
1
340
GLM-5.3-Flash from @zai_org is on Zyphra Inference, served with the Zyphra Optimization Engine (ZOE) on @AMD Instinct MI300X.
With ZOE, p90 generation is 3.1x faster per user and p90 time to first token is 10.7x lower than stock SGLang.
Try it at: cloud.zyphra.com
Most of the revenue comes from cached input, since about 95% of AgentX input tokens hit the prefix cache.
Zyphra GLM-5.3-Flash pricing per 1M tokens:
Input $0.15
Cached input $0.03
Output $0.25
Model card at zyphra.com/models/glm-5.3-fl…
1
296
@ZyphraAI is a full stack superintelligence company based in San Francisco on a mission to enable sovereign AI through open models and multisilicon infrastructure.
Apply to join us! jobs.ashbyhq.com/zyphra
253