Run DeepSeek-R1 7B, a distilled model built on Qwen-7B, behind an OpenAI-compatible API on a GPU instance in your own AWS account.
DeepSeek-R1 7B is a distilled version of DeepSeek R1 built on the Qwen-7B architecture. It is efficient to run, strong at Chinese-language tasks and capable across other languages.
DeepSeek R1 capabilities in a 7B model.
Optimised for Chinese with strong multilingual support.
Use OpenAI SDKs with your own endpoint.
Browse and test the API at /docs on your server.
Runs on g4dn instance types in your account.
Prompts and outputs stay in your own account.
The AMI ships the model and an OpenAI-compatible API server on Ubuntu 22.04.
Make sure your account has vCPU quota for g4dn instances in your region.
Subscribe on AWS Marketplace, accept the terms and launch the CloudFormation stack.
Create a DNS record for your domain with the server's public IP.
Open /docs on your domain or IP, then point your OpenAI client at your server.
An AMI on AWS Marketplace, launched on a g4dn GPU instance in your own AWS account with a CloudFormation stack.
Launch on AWS The server follows the OpenAI API, so clients only need a new base URL.
| Feature | Supported |
|---|---|
| Chat and text completions | Yes |
| Model listing | Yes |
| Embeddings | No |
| Interactive docs at /docs | Yes |
A distilled DeepSeek R1 model built on the Qwen-7B architecture, optimised for Chinese with strong multilingual ability.
Yes. Browse the API at /docs on your domain or public IP and use OpenAI SDKs against your server. Embeddings are not supported.
g4dn instances, from g4dn.xlarge up to g4dn.metal.
Per hour of software use, which varies by instance type, plus the EC2 instance. A 365-day contract is also available.
Use your key pair with the ubuntu username.
That happens when the domain is not on Route 53. Run the certificate script on the server manually.
DeepSeek, Llama, Qwen and more, compared for self-hosting.
Swap the OpenAI base URL without rewriting app code.
The DeepSeek model built for code.
We build and run self-hosted AI infrastructure every day, from model selection to GPU sizing. Tell us what you need.
Contact us