Rapid Router: Every AI model.
One API key.
Claude, GPT, Gemini and 500+ more through the SDK you already use.
593 models · 63 makers · UPI + GST
OpenAI-compatible. No subscription.
One key, every major lab
One key. Every lab. The right route, every request.
Each request goes to the cheapest healthy provider for that model and fails over on its own. Latency and uptime for every route are measured every hour and published on each model page.
Live prices and latencies. Reply text is illustrative.
- 593
- Live models: 430 chat, 163 image, video, speech and more
- 63
- Model authors
- UPI
- Top-ups in rupees, GST invoice included
- 5%
- Fee, charged once at top-up
GET /v1/models lists 481 chat IDs: the 430 chat models plus 51 provider-pinned routes to them.
The work a gateway should do. Done once, here.
Keys, balances, metering and spend tracking are the same problems for everyone. Solve them here instead of in every service you own.
Switch models by changing one string
Move from Claude to GPT to Gemini without touching anything else. Same request shape, same response shape, same SDK you already use.
Priced in rupees, invoiced with GST
Every model carries a per-million price in the catalog, and a GST invoice follows every recharge. A bill your finance team will actually approve.
One balance behind every model
No separate account, card or quota per provider. Add credit once and spend it across the whole catalog.
See exactly where the money went
Every request is logged with its model, tokens and cost. Spend breaks down by key, and each key can carry its own monthly cap.
Rate limits you can plan against
From 60 requests a minute per key, rising with your balance to 1,200, and from 8 in flight per account up to 64. The count comes back in the response headers. Free models stay at 60 and 8.
Your key is a hash on our side
The secret is shown once at creation and never stored. Lose it and you replace it, which is the only safe way round.
Two lines. That is the migration.
Change the base URL and the model string. No new SDK to learn, nothing to rewrite, no per-provider branching left in your code. The package your snippet imports is only an HTTP client, and every request comes here.
- 1Create an accountSign in with Google, GitHub or an email and password. Your account is ready immediately.
- 2Add creditsTop up with UPI. One balance in rupees, valid against any model or provider.
- 3Ship itIssue a key, set a budget cap, start sending traffic.
curl https://rapidrouter.ai/api/v1/chat/completions \
-H "Authorization: Bearer $RAPIDROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "google/gemini-3.7-flash",
"messages": [{ "role": "user", "content": "Hello" }]
}'The frontier, one key away.
Each lab's top model from the last 90 days, the one it prices highest. Callable right now with your key.
Paying in dollars does not save you the 18% GST. It only moves the paperwork.
The tax is the same either way; what differs is who has to handle it.
See the sum, both waysOur 18% is claimable as input tax credit. Buy abroad and a provider stops charging it once you hand over your GSTIN, which is not a discount: you then owe the same 18% yourself under reverse charge, deposit it, and claim it back.
You pay 18% either way. A foreign provider selling to an unregistered Indian customer has to charge and deposit Indian GST under the OIDAR rules, so going direct changes the invoice, not the tax.
API pricing in India, by model family
Every family below is live through one key. Each page lists the models with their price per million tokens in rupees, context window and what they support.
What is Rapid Router? And other good questions.
- What is Rapid Router?
- Rapid Router is an AI model gateway: one OpenAI-compatible API in front of 500+ live models, Claude, GPT and Gemini among them. You hold one key and one balance, and switch models by changing one string.
- Is this the Rapid Router coding game?
- No. The Code for Life Rapid Router is a free coding game for schools by Ocado Technology. Rapid Router at rapidrouter.ai is an AI API gateway for developers, built and run in India. The two are unrelated.
- How is Rapid Router different from OpenRouter?
- Same idea, same request shape: change the base URL and the key, and most OpenRouter model ids work as they are (each model page shows ours). The differences are rupee billing with UPI and GST invoices, image, video and speech models behind the same key, a batch API, and support that answers in Indian hours.
- What does it cost?
- Every model is billed at provider list price per million tokens, with a platform fee of 5% charged once when you buy credits. No seats, no subscription, no minimum.
- Do I need to change my code?
- No. Set the base URL to https://rapidrouter.ai/api/v1 in the OpenAI SDK or any OpenAI-compatible client, or https://rapidrouter.ai/api in an Anthropic SDK (it adds /v1/messages itself), and use your Rapid Router key. Streaming, tools, images and audio work as before.
- Is there a free tier?
- Yes. Every account gets a daily allowance of free requests on eligible models before any top-up, and early-stage teams can apply to the startup program for starting credits.
Your first request is a minute away.
Provider list price on every request, with a fee of 5% charged once when you buy credits. Rupee top-ups convert at market plus our currency margin, shown before you pay, with a GST invoice on every recharge. No seats, no minimums.