Infinity Intelligence
The AI division of Infinity Productions
Four generations of models for institutions, enterprises and governments — and a deliberate decision about what comes after them.
What we are
Infinity Intelligence is the artificial intelligence research and systems division of Infinity Productions. We build machine learning models, the infrastructure to train and run them, and the intelligent systems that sit on top — for institutions, enterprises and governments rather than for a consumer market.
That focus shaped every technical decision we made. Closed weights, on-premise and air-gapped operation, and a strong bias toward saying nothing over saying something wrong are not features we added late; they are the reason the division existed.
In June 2026 we concluded our frontier model program. Sustaining state-of-the-art training at our size stopped being viable, so we shut those efforts down deliberately rather than let them decay. Infinity Intelligence Labs is now the centre of the division, and our work continues through PolyCode and the WillMe GPT models.
How we got here
First generation
Development of the first Xora models runs from 2021 through to release in 2023, establishing the division’s focus on institutional rather than consumer deployment.
Second generation
The second generation is released at the end of 2024. The Xora family finds its footing in deployments where closed weights and on-premise operation are requirements rather than preferences.
Third generation
The third generation arrives in the first quarter of 2025. Xora 3.5 becomes the most widely deployed model the division produces, and work begins on Sentra and on the diffusion approach that becomes PolyCode.
Fourth generation, and the end of the program
In June the division concludes its frontier model program, cancels Sentra, discontinues PySML, and moves its priority to Labs, PolyCode and the WillMe GPT models. Xora 4 Flash reaches full release in August 2026 as the last model in the family.
The wider ecosystem
Infinity Intelligence is one part of a group of connected projects.
Infinity Productions
The parent organization, overseeing game development, technology research and the integration of every product and platform in the group.
WillMe AI
Our public-facing platform and experimental laboratory, where architectures like the linear RNNs behind WillMe GPT are tested with real users.
RAID Chat
A privacy-centred communication platform built to put control back in the hands of its users, with no back door for anyone — government or otherwise.
Infinity Intelligence Labs
Our research subdivision, now the centre of the division: experimental architectures, efficiency work, and the two model programs that continue.
Working with us
We evaluate requests for bespoke model development and partner with a small number of organizations on architectures built from the ground up. If a project needs capability beyond off-the-shelf systems, or has operational and data constraints that rule out a hosted service, we would like to hear about it.
Start a conversation →