# Xinference

> Australian-owned AI inference platform. Deploy 300+ models on managed cloud, private infrastructure, or fully air-gapped environments. One OpenAI-compatible API. You choose the inference engine — vLLM, SGLang, MLX, or Transformers.

Xinference is built end-to-end for one job: serving large AI models at production scale. The same runtime that powers managed cloud, private deployments, and air-gapped clusters. NVIDIA-optimised. Up to 2× the throughput of standard open-source stacks on equivalent hardware (internal benchmark).

## Company

- **Legal entity:** Xinference Australia Pty Ltd
- **Ownership:** Australian-owned and operated
- **Headquarters:** Australia
- **Team:** Singapore + across the Asia–Pacific region
- **Open-source core:** Apache-2.0 ([github.com/xorbitsai/inference](https://github.com/xorbitsai/inference))

## What Xinference does

- **Dedicated Inference** — single-tenant clusters with up to 2× throughput vs. open source, multiple models per GPU, RBAC, audit logs, SSO/SAML.
- **Model APIs** (launching 18 June 2026) — serverless, token-priced access to popular open and proprietary models. Hosted on Xinference Cloud.
- **Open-Source** — Apache-2.0 inference engine. 300+ models. vLLM, SGLang, Transformers, MLX. Runs on your laptop, GPU cluster, or air-gapped on-prem.

## Deployment options

- **Public Cloud** — deploy into your own AWS, GCP, Azure, Oracle, or Alibaba account. Single-tenant, your VPC, BYO keys.
- **Xinference Cloud** — fully managed by us, usage- or token-priced. Launching 18 June 2026.
- **Self Hosted** — your environment, any infrastructure. Air-gapped option for environments that cannot phone home.

## Models supported

300+ production-ready models across LLM, embedding, reranking, speech, vision, and multimodal. Notable: Llama, Mistral, Gemma, GPT-OSS, DeepSeek, Phi, Qwen, FLUX, Whisper. Custom model registration supported.

## Key URLs

- Homepage: https://xinference.co/
- Product: https://xinference.co/xinference/product.html
- Platform: https://xinference.co/xinference/platform.html
- Customers: https://xinference.co/xinference/customers.html
- About: https://xinference.co/xinference/about.html
- Developers: https://xinference.co/xinference/developers.html
- Solutions: https://xinference.co/xinference/solutions.html
- Documentation: https://inference.readthedocs.io/en/latest/
- GitHub: https://github.com/xorbitsai/inference

## Contact

- Sales / engineering: https://xinference.co/xinference/demo.html
- Email: info@xinference.co