Australian-owned.
Globally trusted.
We build the control plane for sovereign AI inference: run any model, on any infrastructure, without losing control of your data.
Enterprise AI. Full control.
Every prompt sent to a closed API leaves your infrastructure. For regulated industries, that isn't a trade-off, it's a blocker.
Xinference started because ML infrastructure engineers were frustrated with the alternative: rent a black box, pay per token, and lose visibility into where sensitive data goes. We built a unified inference platform that runs any model, on any hardware, inside your own cloud or on-premises environment.
Today Xinference is Australian-owned with a global footprint. Our team of ML infrastructure engineers and researchers is based in Australia and Singapore, building software hardened for real-world enterprise deployment.
The scale we run at today.
Figures we can stand behind, not projections.
*Estimate at list price versus closed-source API list pricing. Actual savings vary by workload and volume.
The principles behind our work.
Sovereignty by design
Your models and your data run inside your own infrastructure. Nothing leaves without your say so.
Engineering rigor
Built by ML infrastructure engineers for production workloads, not a research demo.
Straightforward pricing
Clear, list-price economics. No hidden egress fees, no black-box billing.
Embedded support
Enterprise accounts work directly with engineers who know their deployment, not a ticket queue.
Built by an infrastructure engineer, for infrastructure teams.
Chris Qin is the founder and CEO of Xinference. He created and led Mars, an open-source framework for large-scale distributed data computation, and has spent his career building systems that run heavy compute reliably at scale.
He started Xinference on a simple conviction: as AI moves into regulated, high-stakes work, enterprises cannot hand their data and their models to someone else's black box. Inference should run on infrastructure the customer controls, with the performance and economics of a purpose-built platform.
That is what Xinference builds today: any open model, on your own hardware, through one OpenAI-compatible API.
“Your models and your data should never have to leave your environment for AI to work.”
Xinference pools heterogeneous GPU resources in our private cloud, significantly reducing infrastructure cost for 6,000+ users while turning the latest open-source models into core productivity.
Join the team.
Build the infrastructure enterprise AI runs on.
Work on production infrastructure that regulated enterprises depend on every day, in Australia and Singapore.
We are hiring across engineering and field teams.
Run inference
on your terms.
Any model. Any infrastructure. Complete control from day one.