/01
Price is only half the story.
Fresh canary prompts, model identity checks, and usage audits put every offer through its paces. See the evidence before you route.
Inside verification↗ THE VERIFIED INFERENCE EXCHANGE
More intelligence.
Less weight.
Powerful models. Competitive prices.
One beautifully simple API.
Connect to an open market of inference capacity.
We find the right route. You build what comes next.
⌘ OpenAI & Anthropic compatible Your stack, upgraded.
TC / EXCHANGE CORE
BUILT TO DO MORE. WITH LESS.+
THE THIN LAYERBetween your idea and every answer.
YOUR EXISTING STACK
⌘ OpenAI SDK
Anthropic
▧ vLLM
⟨/⟩ SGLang
One integration.More possibilities ↗
01 / INTELLIGENT BY DESIGN
One request.
The right route.
Don’t settle for a single provider.
Put an entire market to work for you.
You set the rules. We find the fit.
Less spend. Same ambition.
Choose the lowest-priced healthy offer that meets your model and verification requirements.
REQUEST ROUTINGILLUSTRATION / NOT LIVE DATA
↗Your app
ⅡThinCompute
Offer AEligible
Offer BEligible
Offer C✓ Selected
✓Model matches✓Verification passed✓Lowest eligible price
YOUR POLICY
/p/cheap+verified/v102 / TRUST IS PART OF THE STACK
Less black box.
More peace of mind.
/02
A paper trail. Without the paper.
Every completed call comes with a signed receipt. The model, the seller, the tokens, the cost. Yours to inspect and independently verify.
Explore signed receipts↗/03
An answer worth finishing.
If a stream drops, rescue can continue the response through another eligible offer. Optional hedging gives slow first responses a second route.
How resilience works↗TRANSPARENCY, BY DEFAULT.
Behavioral verification reduces risk. It is not hardware attestation. Understand the checks ↗
03 / A SMALL CHANGE. A BIG OPENING.
Keep your code.
Change the
possibilities.
Same familiar SDK. A new world of capacity.
Change your endpoint, add your key, and
get back to the interesting part.
✓ OpenAI-compatible✓ Streaming & tool calls✓ Anthropic Messages
from openai import OpenAIimport os client = OpenAI( base_url="http://localhost:8787/v1", api_key=os.environ["THIN_API_KEY"],) response = client.chat.completions.create( model="YOUR_LISTED_MODEL", messages=[{"role": "user", "content": "Hello, world."}],)Set your API key and choose a model from the market.
01What exactly is ThinCompute?+
An open exchange for AI inference. Sellers list available model capacity; your requests route to eligible offers according to your price, reliability, and privacy requirements. You use one API key and an OpenAI-compatible endpoint.
02Do I need to change my application?+
If your application uses an OpenAI-compatible SDK, start by updating the base URL and API key, then choose a listed model. An Anthropic Messages endpoint is also available. Check the docs for supported parameters and model-specific differences.
03What does verified mean?+
An offer has passed behavioral checks covering model identity, known-answer prompts, instruction following, usage reporting, and context recall. These checks reduce risk; they do not cryptographically prove the model weights. Hardware attestation is planned.
04How does pricing work?+
Sellers set input and output prices. Buyers pay the listed rates; the exchange retains a 5% seller fee. Prices and availability vary by offer. You can set price ceilings, budgets, and minimum-discount rules before routing.
05Can I sell my unused capacity?+
Yes, if you operate an OpenAI-compatible endpoint and have the right to sell its capacity. Connect your endpoint, set your prices and availability, and complete verification. Visit Sell capacity for the onboarding guide.
LESS BETWEEN YOU AND WHAT’S NEXT