Hosted in Australia
Sovereign inference.
Run any model. One API.
Kimi K3 and DeepSeek through one OpenAI-compatible API, hosted in Australia. Sign up, get a key and send your first request in minutes. Pay per token.
Already have an account? Log in
client = OpenAI(
base_url="https://api.xinference.co/v1",
api_key="XINFERENCE_API_KEY",
)
# two lines. that's the migration.Trusted in production by
Model API pricing
Pay per token.
Same per-token price as the model vendor, hosted in Australia. Create an account, get your key and pay only for the tokens you use.
Served from dedicated GPUs in Australia.
| Model | What it’s for | Input | Output |
|---|---|---|---|
 Kimi K3 | Frontier reasoning and agentic workloads | $3.00 | $15.00 |
 DeepSeek V4.1 Flash | Long-context work and RAG | $0.30 | $1.20 |
| DeepSeek V4.1 Flash costs up to 53% less than DeepSeek on Amazon Bedrock in Sydney. It costs up to 88% less per token than Claude Sonnet 5.5. List prices as of October 2026. | |||
| USD per 1M tokens. Don’t see the model you need? Talk to us about adding it. | |||
Why teams switch
Most teams start on the Model API and move to dedicated compute as usage grows.
See customer results→No lock-in
Open models behind one OpenAI-compatible API. Swap models, hardware or clouds without rebuilding your applications.
Cost certainty
Published prices per token. Pay for what you use, and move to a set cost per GPU hour as you grow.
Hosted in Australia
Runs in Australian data centres. No training on your data. No prompt or response content kept by default.
Start in minutes
Sign up, create a key and change two lines of code. No sales call needed.
In production.
Quick-service restaurantsOne platform behind 50+ AI use cases.
Yum! runs 50+ AI use cases across KFC, Pizza Hut and Taco Bell, at 1.52M requests a day across dual data centres.
Financial services6,000-user AI on one pooled cluster.
360,000 requests per day on 72 pooled GPUs, serving 6,000 users.
InsuranceA regulated AI platform, compliant by design.
500K+ daily requests, replacing a self-managed vLLM and Kubernetes stack.
What you can build on Xinference.
Most production workloads already run on open models, with the same or similar results as closed-source APIs.
Search and RAG
Answer questions over your own documents.
Conversational AI
Customer assistants and internal helpdesks.
Agents and function calling
Multi-step reasoning and tool use, on frontier open models such as Kimi K3 and DeepSeek V4.1 Flash. You can also build them with Xagent, which runs agents on your own models.
Coding assistance
IDE copilots and code generation, on long-context models served with streaming.
Summarisation and extraction
Document parsing, summarisation and classification, including high-volume batch pipelines.
Long-context work
Contracts, reports and codebases in a single request.