Xorbits
Xinference (Xorbits Inference) is an Australian-owned AI inference company and open-source platform for deploying, serving, and integrating open-source language, speech, embedding, and multimodal models. Its commercial platform provides a unified OpenAI-compatible API, managed inference, and private/on-premises deployment so enterprises can control their data, models, infrastructure, and costs.
At a Glance
- Financial services
- Insurance
- Healthcare
- Government and public sector
- +4 more
AI Tools by Xorbits
(1)
Xinference
Open Source Model Serving API
Discussions
No discussions yet
Be the first to start a discussion about Xorbits
Latest News
Published a guide on deploying Xinference as a self-hosted OpenAI API alternative.
Recap of The AI Summit Australia 2026 in Melbourne and Xinference's participation.
Recap of the AI & Data Governance Summit 2026, including a panel moderated by Paul Towers.
Announced an Asia-Pacific partnership with BioQuest Advisory for enterprise AI delivery.
Products & Services
Open-source model-serving library/platform for deploying and serving LLMs, speech-recognition, embedding, rerank, image, audio, video, and other multimodal models locally, in the cloud, or on premises.
Commercial edition for production teams: a control plane for model repositories, deployment, routing, GPU pooling, observability, governance, SSO, audit logs, and private deployment in a customer's cloud or data centre.
Hosted Australian-region API for 300+ open models through an OpenAI-compatible endpoint, billed per token and supporting streaming, function calling, structured output, embeddings, and vision models.
Managed dedicated GPU inference or deployment on the customer's own cloud, data centre, or fully offline infrastructure, using the same API and model catalogue.
Market Position
Xinference positions itself as a sovereign, open-model control plane between single-model servers and closed managed APIs: one OpenAI-compatible interface across 300+ models, engines, GPUs, clouds, and private infrastructure. It emphasizes Australian/in-customer data residency, portability and no lock-in, governance, and lower cost than closed APIs. The relevant alternatives include Ollama, vLLM, SGLang, NVIDIA NIM, Ray Serve, and BentoML; Xinference differentiates by combining broad model coverage, multi-engine serving, lifecycle operations, enterprise controls, and managed plus self-hosted deployment.
Leadership
Founders
Chris Qin (Qin Xuye)
Founder and CEO of Xinference/XorbitsAI; former BDFL and core developer of the Mars open-source project, which he built at Alibaba Cloud to run pandas and NumPy workloads reliably at scale.
Executive Team
Chris Qin (Qin Xuye)
Founder and CEO
Former BDFL and core developer of Mars; previously built scalable data-science infrastructure at Alibaba Cloud.
Founding Story
Chris Qin started Xinference in 2023 after repeatedly encountering the problem of making code that worked on one machine run reliably on a thousand. His experience building Mars at Alibaba Cloud led him to apply the same operational insight to open models: teams needed a simple way to deploy and manage models on their own hardware rather than deal with a fragmented inference stack or a closed provider's black box.
Business Model
Revenue Model
Usage-based hosted Model API billing per token; managed dedicated GPU/cloud deployment priced by dedicated compute; enterprise/private deployment and support sold through contact-sales agreements; and a Xinference plus Xagent bundle priced for hosted users.
Pricing Tiers
Hosted open models in Australia, including GLM 5.3, DeepSeek 4.1 Flash, and Kimi K3.
Managed dedicated H200, H100, B200, A100, or RTX6000 Pro instances; volume discounts available.
Customer cloud, data centre, or fully offline deployment with dedicated support.
Hosted in Australia; private deployment plus the Xagent agent platform.
Target Markets
- Financial services
- Insurance
- Healthcare
- Government and public sector
- Legal
- Education and research
- Enterprise RAG and document search
- Conversational AI and customer/internal helpdesks
- Agents, function calling, and tool use
- Coding assistance and IDE copilots
- Summarization, extraction, document parsing, and meeting minutes
- Financial-services investment research, compliance, KYC, and contact centres
- Siemens
- Yum! Brands
- AIA
- TFC OpticalComms