planned, not live

Private inference on CPUs.

Your prompts are sealed from the host, inside AMD SEV-SNP or Intel TDX. You can verify the attestation before you send data. Planned, not live.

Pricing: TBD, set after we measure our own hardware.

try it

The API example on the home page. No account, no card.

Preview

Open models on CPUs
No API key
Not live yet

See the example

api access

Open models on CPUs, sealed from the host, with an OpenAI-compatible API.

TBD

Prompts sealed from the host
Attestation you can verify
AMD SEV-SNP or Intel TDX
OpenAI-compatible API
Your own API keys
Planned, not live

Create an account

more than that

Higher volume, or a model we do not serve yet.

Talk to us

Volume pricing
Model requests
Direct line to the people who run it

Email us

developer-first

Built for developers

Drop in an OpenAI-compatible API. The planned flow is to verify an attestation report first, then send the request encrypted to the enclave. This is not live yet.

OpenAI-compatible /v1/chat/completions endpoint
Planned attestation at GET /v1/attestation
API key authentication
Inference on CPUs, sealed from the host

terminal

curl https://api.cpuinference.ai/v1/chat/completions \
  -H "Authorization: Bearer $CPU_INFERENCE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "lfm3-1.7b",
    "messages": [
      {"role": "user", "content": "Hello!"}
    ]
  }'

Preview: the API host above is a placeholder and is not live.

Ready to start?

Create an account and generate an API key. Inference is not live yet.

Create an account

Want to try it first? Use the example on the home page or read the docs. Questions? Get in touch.