Investment research
Evaluate document analysis and research synthesis against your own acceptance criteria. Compare output quality, latency and total token cost.
Get secure, scalable frontier AI.
Compare models, pricing and privacy for financial, legal and enterprise teams.
Developer quick start
One API.
Confidential tokens.
Use your TrustedRouter API key and credits with the OpenAI-compatible API. Keep the same SDK, request arguments and streaming responses.
- Create an API key and set
TR_API_KEYin your environment. - Set your client's base URL to
https://api.confidential.nytokenexchange.com/v1. - Choose
nyte/autoand requireprovider.min_privacy: "confidential".
Confidential inference
Require verified confidential compute, provider-side encryption and zero data retention on every eligible route.
curl --no-buffer --fail-with-body https://api.confidential.nytokenexchange.com/v1/chat/completions \
-H "Authorization: Bearer $TR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "nyte/auto",
"messages": [{"role": "user", "content": "Hello"}],
"provider": {"min_privacy": "confidential"},
"max_tokens": 1024,
"stream": true
}'Confidential requests fail closed when no route meets these requirements. The standard endpoint, https://api.nytokenexchange.com/v1, also offers the broader catalog, with privacy determined by your selected route.
nyte/auto, nyte/fast, nyte/zdr and every other nyte/ model name accept the same arguments as their trustedrouter/ equivalents. Your account, balance and key limits stay the same. Explore models.
See the exchange in action
One gateway.
Smarter model routing.
Watch a 90-second walkthrough of routing different workloads through one gateway.
Illustrative demo. Model availability, pricing and savings vary.
Workload priorities
Verifiably secure compute for frontier industries.
Legal operations
Review retention policies and provider eligibility before routing client material. Test document workflows with a controlled evaluation set.
Financial platforms
Define approved routes, fallback choices and application spending limits before putting an AI feature into production.
How the exchange works
Buy capacity.
Choose how it is served.
Compare models and provider policies through one API. Confirm processing locations and route requirements before you buy.
Plan an enterprise pilot- Plan capacity for your workload
- Route through pre-approved providers
- Review attestation evidence
NYTE requires confidential AI routes. Show your customers the NYTE Privacy Secured seal.
Take it to your team
A clearer way to buy AI.
Secure. Intelligent. Cheaper. Share the 13-page buyer deck with your engineering, security and finance teams. See how the exchange brings more providers into competition for your workloads, how confidential routes are verified, and how to start a 30-day pilot.
- One endpoint, more suppliers and competitive pricing
- Combination models and verified confidential routes
- The team, customer results and a 30-day pilot
Prefer a conversation? enterprise@trustedrouter.com
A market needs trust.
TrustedRouter does not keep prompt or output logs. Zero data retention is a provider policy, separate from confidential inference. Review provider retention and attestation evidence for your route.
Review the trust centerConfidential models
Tinfoil · Confidential + E2EE
USD / 1M tokens · input / output
Attestation
For suppliers
Inference supplier?
Bring it to market.
Offer inference capacity for research, document and enterprise workloads. Share throughput, pricing and privacy evidence so buyers can evaluate your service.
Show your service
Models, pricing, capacity and the regions you serve. Credentials are shared only through the secure onboarding channel.
Make it verifiable
Conformance tests against the public provider suite, retention policy and operational evidence.
Serve real workloads
Approved endpoints can be matched with buyer requirements. Listing and purchase volumes depend on review and demand.
Open source for suppliers
Your GPUs. Your brand.
A market for spare inference.
Offer spare GPU capacity as white-label model inference under your own provider brand. Publish your models, capacity and rates, then let buyers route eligible workloads to your service.
The inference stack is open source.
We give operators Apache-2.0 software for building a confidential neocloud: an OpenAI-compatible serving endpoint, model packaging, hardware attestation and provider conformance tooling. You operate the hardware and inference. TrustedRouter handles customer access, routing and billing.
Pre-alpha. Confidential service qualification requires supported hardware, a verified serving runtime and independent attestation. See the repository for tested components and current readiness.
- 01
Bring spare capacity
Choose the models your hardware can serve. Keep your operator identity and offer inference under your own brand.
- 02
Verify the full stack
Qualify the hardware, runtime, attestation and zero-retention policy before a confidential route goes live.
- 03
List your inference rates
Submit pricing, throughput and available capacity. Marketplace listing and payout terms are agreed during onboarding.
A reference price for intelligence.
The NYTE Token Index shows published inference prices by capability grade. Use it as a reference when pricing spare capacity. Actual provider quotes and commercial terms are agreed separately.
Before you begin
Good questions.
What exactly am I buying?
Access to model inference through TrustedRouter. AI tokens measure model inputs and outputs. This is an inference service, not a securities or cryptocurrency exchange.
Does this market guarantee local processing?
The regional name identifies the audience for this page. Processing location, residency and provider eligibility depend on the route and agreed terms. Confirm these requirements with us before sending sensitive data.
Can we start with one workload?
Yes. Bring a representative, non-sensitive task set and expected volume. We can compare suitable models and providers, then define quality, cost and privacy requirements for a pilot.
I run inference. Can I sell capacity here?
We onboard inference providers with working model endpoints. Share your infrastructure and available capacity so we can discuss integration. Listing and purchase volumes are subject to review and demand.