Best execution
for every token
Hyperouter is a smart order router for AI. Each request is quoted across every provider serving the model and filled at the best price, latency and uptime.
Markets
List prices per 1M tokens, set by each model’s developer. Hyperouter charges exactly the same.
| Model | Input / 1M | Output / 1M | Typical call |
|---|
Prices shown at time of publishing. The Models API returns live prices.
How an order gets filled
Popular models trade on several venues. Hyperouter quotes all of them on every request.
- Quote. Price, latency and recent error rate are checked for every provider serving the model.
- Route. The order goes to the best venue within your constraints. Ties on price go to the fastest.
- Fail over. If a venue errors or stalls, the order moves to the next one. Failed attempts are not billed.
- Settle. Cost is debited from your balance at the venue’s price, with a line item for every call.
anthropic/claude-sonnet-5Venue quotes| Venue | In / Out | Latency | |
|---|---|---|---|
| Anthropic | $3.00 / $15.00 | 0.41s | Filled |
| Amazon Bedrock | $3.00 / $15.00 | 0.52s | |
| Google Vertex | $3.00 / $15.00 | 0.61s | |
| Venue D | $3.00 / $15.00 | — | Halted |
Same price on three venues, so the order fills at the lowest latency. Venue D is skipped after elevated error rates in the last 60 seconds.
Works with the OpenAI SDK
Swap the base URL and key. Streaming, tool calls, JSON mode and image input behave the same on every venue.
import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://hyperouter.app/api/v1", apiKey: process.env.HYPEROUTER_API_KEY, }); const res = await client.chat.completions.create({ model: "anthropic/claude-sonnet-5", messages: [{ role: "user", content: "Hello" }], });
import os from openai import OpenAI client = OpenAI( base_url="https://hyperouter.app/api/v1", api_key=os.environ["HYPEROUTER_API_KEY"], ) res = client.chat.completions.create( model="google/gemini-3-pro", messages=[{"role": "user", "content": "Hello"}], )
# provider.sort and models[] control routing and fallback curl https://hyperouter.app/api/v1/chat/completions \ -H "Authorization: Bearer $HYPEROUTER_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "models": ["openai/gpt-5", "anthropic/claude-sonnet-5"], "provider": { "sort": "price" }, "messages": [{"role": "user", "content": "Hello"}] }'
Model access should work like a market: open to anyone, priced in the open, and routed to whoever serves it best.
No markup
Tokens are billed at the provider’s list price. Hyperouter earns a fee on top-ups, not a spread on your usage.
Transparent routing
Every response reports the venue it was filled at and what it cost. The same data is in your usage export.
Your data stays yours
Prompts are not logged by default. Restrict routing to zero-retention venues with one setting. See the Privacy Policy.
Pricing
No subscriptions, no minimums. Top up a balance and pay for the tokens you use at the provider’s price.
FAQ
What is Hyperouter?
One API that routes your requests to model providers such as OpenAI, Anthropic, Google, DeepSeek, xAI, Alibaba Qwen, Moonshot and Zhipu. One key, one balance, one statement.
How do I get access?
Hyperouter is onboarding teams ahead of public launch. Request an API key and we’ll reply from account@hyperouter.app.
Is it more expensive than going direct?
No. Token prices match the provider’s list price. The only fee is 3.0% when you top up by card, or 2.5% with USDC or USDT.
What happens when a provider goes down?
The request moves to another venue serving the same model, or to the next model in your models list. Failed attempts are not billed.
Do you store my prompts?
Not by default. Logging is off unless you enable it, and you can restrict routing to venues that retain no data. Details in the Privacy Policy.
Can I keep my existing code?
Yes. Set the base URL to https://hyperouter.app/api/v1 and use your Hyperouter key. Anything built for the OpenAI Chat Completions API works.