Inventing the
intelligence that
invents tomorrow.
AI999X is a research lab building self-improving intelligence — and releasing open models across every field of AI, from language and vision to audio, robotics and science.
Intelligence that
improves itself.
Every generation of models we build helps build the next one — faster, smaller and more capable. We believe recursive self-improvement is the shortest path to intelligence that can advance science, engineering and everyday life.
Self-improving
Systems that learn from their own results and grow more capable with every cycle.
Efficient
Compact models that run anywhere — on a laptop, a phone, in a browser — with no cloud required.
Open
Our models are released openly, so anyone can use them, study them and build on them.
One lab.
Every kind of model.
We are building toward the full range of machine learning — one release at a time, each measured before it ships.
Language
Understanding, reasoning, translation and generation.
Vision
Seeing, detecting, segmenting and creating images and video.
Audio & speech
Recognising, understanding and producing sound and voice.
Multimodal
Models that read, see and listen at once.
Tabular & time series
Prediction and forecasting over structured data.
Robotics & agents
Policies that plan and act in the world.
Science & mathematics
Physics, mathematics and the sciences, solved step by step.
Graphs & beyond
Molecules, networks, knowledge — and fields still to come.
Fresh from the lab.
The newest AI999X releases, live from Hugging Face.
“The last invention humanity needs to make is an intelligence that keeps inventing.”
— The idea AI999X is built on
Follow the work.
New models arrive often. Follow AI999X on Hugging Face and GitHub, or write to us.