Azure OpenAI Alternative — Attested Private Inference
Keep the privacy that put you on Azure OpenAI. Hardware-attested inference, 220+ models, automatic provider fallback, OpenAI-compatible API. Verify it live.
Azure OpenAI promises privacy in a contract; TrustedRouter proves it with hardware attestation you can check live.
Teams comparing alternatives to Azure OpenAI usually landed there for one reason: the data-handling terms. Then the operational bill arrives. Deployment quotas that take support tickets to raise. Models retired on Microsoft's schedule, with the migration work left to you. A catalog that only ever holds one vendor's models.
TrustedRouter keeps the property you were buying. Every prompt runs through a hardware-attested confidential-compute gateway. The gateway source is open, the image digest is published, and one curl returns an attestation signed by the CPU vendor's root key, bound to your live TLS session. TrustedRouter never logs prompt or output content. Ordinary synchronous and streaming inference does not retain it. The opt-in Batch API temporarily retains enclave-encrypted artifacts for up to 30 days. Privacy you can check beats privacy you're promised.
from openai import OpenAI
client = OpenAI(
base_url="https://api.trustedrouter.com/v1", # was: your Azure endpoint
api_key="tr-your-key",
)
reply = client.chat.completions.create(
model="trustedrouter/auto",
messages=[{"role": "user", "content": "Hello"}],
)Automatic provider rollover
Routes are served across 30+ providers. When a provider goes down, requests roll over to another serving the route automatically. Latency numbers come from real routed requests, published at the latency benchmarks.
220+ models, 30+ providers
Claude, GPT, Gemini, DeepSeek V4, Kimi K2.7, Qwen 3.5, GLM-5 and more behind one OpenAI-compatible API. Meta-routes decide per request: trustedrouter/auto for best fit, trustedrouter/cheap for the cheapest capable route. trustedrouter/frontier, trustedrouter/advanced, trustedrouter/professional, and trustedrouter/efficient use fixed ordered pools with no synthesis or implicit privacy floor. Compare at /choose.
Set privacy per route
Three tiers: Open, Zero-Data-Retention, and TEE, where even the serving provider cannot read the prompt. With Confidential privacy (provider.min_privacy "confidential"), prompts never reach the model vendor: a Confidential request routes only to third-party confidential-compute hosts, and fails rather than fall back to the vendor's own API. Details on the security page.
Verify, don't trust.
TrustedRouter publishes a DPA, a BAA, a subprocessor list, and SOC 2 and HIPAA readiness documentation. None of these claim a certification. They give your compliance team a verifiable architecture to assess directly, with the live attestation as evidence. An EU gateway with EU-focused routing covers European workloads.
The pattern holds at production scale: a document-processing company scaled from zero to over a billion tokens per week through the attested gateway within three weeks of onboarding. Pricing is a thin markup over provider list prices, per model, at /pricing. Try any route first in the playground.
Current routes, prices, privacy, and measured performance.
Catalog facts come from the routes currently configured in TrustedRouter. Performance uses the same cached metadata snapshot as the public leaderboard. Prompts and outputs are not part of these measurements.
| Model | Providers | Context | Input | Output | Privacy | Measured route |
|---|---|---|---|---|---|---|
Anthropic: Claude Opus 5.5anthropic/claude-opus-5.5 |
|
1,000,000 | $4.22/1M to $4.642/1M | $21.1/1M to $23.21/1M | ZDR | 2101 ms TTFT amazon-bedrock · 100.00% available · n=39 |
Anthropic: Claude Opus 4.8anthropic/claude-opus-4.8 |
|
1,000,000 | $5.275/1M to $5.8025/1M | $26.375/1M to $29.0125/1M | ZDR 5 cited scores | 1816 ms TTFT amazon-bedrock · 80 tok/s · 100.00% available · n=21 |
Google: Gemini 2.5 Flashgoogle/gemini-2.5-flash |
|
1,048,576 | $0.3165/1M | $2.6375/1M | ZDR | 1200 ms TTFT google-ai-studio · 12.50% available · n=10 |
Google: Gemini 3.1 Pro Previewgoogle/gemini-3.1-pro-preview |
|
1,048,576 | $2.11/1M | $12.66/1M | varies 3 cited scores | measured atlas-cloud · 37 tok/s · 100.00% available · n=10 |
Google: Gemini 3.8 Flashgoogle/gemini-3.8-flash |
|
1,048,576 | $0.79125/1M | $3.95625/1M | varies | 3673 ms TTFT google-ai-studio · 100.00% available · n=10 |
Google: Gemini 3.6 Flashgoogle/gemini-3.6-flash |
|
1,048,576 | $0.79125/1M to $1.5825/1M | $3.95625/1M to $7.9125/1M | varies | 1174 ms TTFT google-ai-studio · 100.00% available · n=8 |
Browse every modelReview provider policiesOpen the full leaderboardSnapshot 2026-10-31T09:17:34.899Z
Questions
Is TrustedRouter SOC 2 or HIPAA certified?
No. TrustedRouter publishes SOC 2 readiness and HIPAA readiness documentation, plus a DPA, BAA, and subprocessor list. None of them claim a certification. The offer is verifiable architecture plus readiness documentation your compliance team can assess. The prompt path runs in a hardware-attested, open-source gateway whose attestation you can check live, which is evidence an audit report alone does not give you.
How do I migrate from Azure OpenAI to TrustedRouter?
Change the base URL. TrustedRouter exposes an OpenAI-compatible API at https://api.trustedrouter.com/v1, so you keep your SDK, your model ids, and your existing code. After the switch you get 220+ model routes across 30+ providers, automatic provider fallback when one goes down, and meta-routes like trustedrouter/auto that pick the best-fit route per request.
How do I verify that prompts stay private?
Request an attestation from the live gateway: call the attestation endpoint with a nonce and you get back a JWT signed by the CPU vendor's root key. The image digest matches the published open-source artifact, and the attestation binds to the live TLS session, so it cannot be replayed. TrustedRouter never logs prompt or output content. Ordinary synchronous and streaming inference does not retain it. The opt-in Batch API temporarily retains enclave-encrypted artifacts for up to 30 days. The full procedure is on the security page.
How does TrustedRouter handle capacity and quotas?
Routes are served across multiple providers. When a provider goes down, requests roll over to another provider serving the route automatically. Latency is measured from real routed requests and published on the latency benchmarks page. Payment is prepaid credits through Stripe, bring-your-own provider keys, or crypto pay-per-request through x402.