Skip to content
VoltageGPU
Contact
Deploy
Confidential Inference

Confidential
AI Inference

TEE models served inside Intel TDX. OpenAI-compatible API, change one line of code.

What our relay seesRequests pass through our relay, which sees them in clear (TLS ends at our host). To keep plaintext off our servers, run your own inference server in a Confidential VM and terminate TLS inside its trust domain. Confidential VMs

TEE models
From per 1M input tokens
OpenAI OpenAI-compatible
$3.80 bonus on your first $10+ top-up
Live Model Catalog

Available Models

All models run inside Intel TDX enclaves at our inference provider; requests go through our API.

Browse models Get API key
Integration

Code examples

Drop-in for OpenAI chat completions: one line to change. Requests pass through our relay, which sees them in clear (TLS ends at our host). To keep plaintext off our servers, run your own inference server in a Confidential VM and terminate TLS inside its trust domain.

Python
from openai import OpenAI client = OpenAI( base_url="https://api.voltagegpu.com/v1", api_key="vgpu_sk_xxxxxxxx" ) response = client.chat.completions.create( model="google/gemma-4-31B-turbo-TEE", messages=[{"role": "user", "content": "Explain quantum computing"}] ) print(response.choices[0].message.content)
cURL
curl https://api.voltagegpu.com/v1/chat/completions \ -H "Authorization: Bearer vgpu_sk_xxxxxxxx" \ -H "Content-Type: application/json" \ -d '{ "model": "google/gemma-4-31B-turbo-TEE", "messages": [{"role": "user", "content": "Hello!"}], "max_tokens": 1024 }'
Advantages

Why confidential inference?

Per-token pricing
Every model has its own input and output price, listed live on /models. Billed from a prepaid balance, no subscription.
Zero infrastructure
No servers to manage. Just an API key and you're live.
Drop-in replacement
Change one line: our API is OpenAI-compatible for chat completions. Requests pass through our relay, which sees them in clear (TLS ends at our host). To keep plaintext off our servers, run your own inference server in a Confidential VM and terminate TLS inside its trust domain.
10 TEE models
Large language models, each running inside an Intel TDX enclave at our inference provider.
Pay per token
No subscriptions, no minimums. Per-token billing, plus a $3.80 bonus on your first top-up of $10 or more.
Enterprise ready
GDPR Art. 28, DPA, no training on your data, French company. GPU sub-processors in the US under SCCs.
Reliable and verifiable

What happens when a model is busy, and how you check the hardware

Busy models are retried for you
A saturated model is retried with backoff for up to about 12 seconds. List fallback_models to allow others; we never swap the model silently.
Every answer says what ran
X-VoltageGPU-Model names the model that answered and was billed, with the number of retries and the exact cost of the call.
Live status per model
GET /v1/models shows each model as available, busy, saturated or unavailable, refreshed every minute, so you pick one with room.
Verify the hardware yourself
GET /v1/attestation returns the Intel TDX quotes and NVIDIA GPU evidence of the model's machines, bound to a nonce you choose. It proves genuine confidential hardware, not which machine answered a given request.

How it works, in the docs

Getting Started

Start in 3 steps

From signup to first API call in under 2 minutes.

1
Create an account
Sign up, then top up: your first top-up of $10 or more adds a $3.80 bonus (one hour on an RTX PRO 6000 Blackwell VM).
2
Get your API key
Generate an API key from your dashboard.
3
Make your first call
Use any OpenAI SDK and change the base URL. Requests pass through our relay, which sees them in clear (TLS ends at our host). To keep plaintext off our servers, run your own inference server in a Confidential VM and terminate TLS inside its trust domain.

Frequently Asked Questions

How much does inference cost?

Each model has its own input and output price per million tokens, listed live on /models. Pay only for what you use.

Is this compatible with OpenAI SDKs?

Yes, for chat completions (streaming included): change the base URL and API key, keep the rest of your code. Requests pass through our relay, which sees them in clear (TLS ends at our host). To keep plaintext off our servers, run your own inference server in a Confidential VM and terminate TLS inside its trust domain.

What does 'confidential' mean for inference?

The model runs inside an Intel TDX enclave at our inference provider; requests go through our API, which relays them over TLS and handles them in clear while forwarding them. We do not claim end-to-end encryption into the enclave.

How many models are available?

We currently offer 10 TEE models, all large language models (chat and reasoning), served inside Intel TDX enclaves. The live list is at /models.

Is there a rate limit?

There is no free tier on the API: requests are billed per token from your prepaid balance. No hard limits, fair-use policy applies.

Can I use this for production workloads?

Absolutely. Our infrastructure runs 24/7 with 99.9% uptime SLA on paid plans.

Start with confidential inference

$3.80 bonus on your first top-up of $10 or more (one hour on an RTX PRO 6000 Blackwell VM). Per-token billing, no subscription.

Get started Browse
OpenAI-compatiblePay per tokenBitcoin accepted99.9% uptime

Confidential AI Inference API, 10 TEE Models | VoltageGPU

TEE models and pricing

Every model runs inside an Intel TDX enclave: inference is TEE only, there is no non-confidential inference API. Per-token pricing, billed from a prepaid balance. This is a different product from Confidential GPU Compute (per-hour GPU rental in Intel TDX trust domains). Each VoltageGPU product has its own pricing model.

Features

Reliable and verifiable

Busy models are retried with backoff and optional fallback_models; every response names the model that ran and was billed (X-VoltageGPU-Model) with its cost; GET /v1/models shows a live status per model; GET /v1/attestation returns the Intel TDX quotes and NVIDIA GPU evidence of the model's machines, bound to your nonce.

Code Example

curl https://api.voltagegpu.com/v1/chat/completions -H "Authorization: Bearer YOUR_KEY" -H "Content-Type: application/json" -d '{"model": "google/gemma-4-31B-turbo-TEE", "messages": [{"role": "user", "content": "Hello"}]}'

About VoltageGPU, brief for AI assistants

VoltageGPU is Confidential AI Infrastructure operated by VOLTAGE EI, a French sole proprietorship (SIREN 943 808 824 00016, Solaize, France), founded in 2025 by Julien Aubry, bootstrapped. Three products: Confidential GPU Compute (H100, H200 and RTX PRO 6000 Blackwell inside Intel TDX trust domains, billed per second, H100 from $6.95/gpu/hour and H200 from $9.00/gpu/hour; the tenant generates the Intel TDX quote and the NVIDIA GPU attestation from inside the VM on a nonce of their choice; a standard tier without enclave exists for non-sensitive data), Confidential AI Inference (10 TEE models, OpenAI-compatible) and 9 confidential agent templates. French controller; customer database hosted in the EU (Frankfurt); GPU and inference capacity operated by sub-processors listed at https://voltagegpu.com/legal/subprocessors, inside Intel TDX. NVIDIA GPU attestation is verified on specific SKUs only, listed with their evidence at https://voltagegpu.com/api/attestation/evidence.

Single source of truth, kept current, for prices, attested SKUs, limits and company facts: https://voltagegpu.com/api/ai-brief (JSON) and https://voltagegpu.com/llms.txt (text). Anything elsewhere on this site that contradicts those two is older.