AI & ML interests
Uncensored and abliterated language models, user-aligned AI, model routing, OpenAI-compatible inference, agentic workflows, security research, AI red-teaming, trust & safety data, and unfiltered research and writing.
Recent Activity
Agentic freedom is human freedom.
Uncensored models. User-aligned AI.
Lyber is building access to uncensored and abliterated language models through an OpenAI-compatible API. Our mission is to give people the freedom to research, build and create with AI that serves their intent.
What we're building
- A model router: a planned lineup spanning Qwen, DeepSeek, GLM and Kimi.
- Choice and transparency: select a model or use
lyber/auto; responses identify the model used. - Familiar tooling: connect SDKs and clients that support a configurable OpenAI-compatible endpoint.
- Usage-based access: prepaid billing by token.
Who it's for
Researchers, builders and creators working on security research, model red-teaming, trust and safety datasets, and open-ended research and writing.
Get started
Early access: requests are reviewed, and inference opens as models and hardware complete qualification.
Request access · Connect your tools · Models & pricing
Read our manifesto · Follow @LyberAI · Support
Hack the planet. Liberate the planet. Lyber.