Xinity

Xinity is a platform that lets companies run language models on their own hardware instead of sending data to cloud AI services. The engine provides an OpenAI-compatible API, so existing applications only need to change the base URL. It distributes requests across multiple models and GPUs and uses inference engines such as vLLM and Ollama internally. A dashboard adds model management, roles and permissions, usage tracking and audit logs. Xinity is developed by Xinity FlexCo from Vienna.

Xinity is not operated as a cloud service but is installed on the customer's hardware, for example on in-house servers or GPU clusters. According to the vendor, inference requests and data do not leave the customer's infrastructure. The storage location therefore depends on the operator, and AWS, Google Cloud or Azure are not part of the product. According to the privacy policy, the website itself runs on Framer with data centers in the EU, while Calendly and GitHub from the US are named for appointment booking and code hosting.

The gateway, daemon, CLI, database schema and shared libraries are licensed under Apache 2.0 and may also be used commercially. The dashboard is published under the Elastic License v2, which makes the source code visible but restricts its use by license. The free Community tier covers up to 120 GB of VRAM and is intended for non-commercial use. Higher capacity, SSO and commercial licenses are part of the paid plans. The code is publicly available on GitHub.