RelayVano API Relay Pricing: How Claude and Gemini Tokens Are Charged

One RelayVano API key, three protocol endpoints

RelayVano is an independent relay made for developers worldwide, and the whole site, dashboard and docs come in six languages: 中文, English, Español, हिन्दी, Português, Français. Prices show in yuan on the Chinese pages and in approximate dollars on the others, so you can read the documentation and check your balance in the language that fits your team. There is no monthly fee: you top up a prepaid balance, and you pay per token as requests come in.

The point of a relay is simple. Instead of juggling several provider accounts and several sets of credentials, you create one API key and send it to https://api.relayvano.com. That single key reaches every model we list, across three familiar endpoints: an OpenAI-compatible /v1, Anthropic /v1/messages, and native Gemini /v1beta. You keep one dashboard, one balance and one usage page no matter which model your prompt is aimed at.

Because RelayVano is built for developers in many markets, the account is also built for different payment habits. New accounts start with $0.40 of credit, which is enough to run a small test and see exactly how your requests are metered before you commit to a larger top-up.

Which client pairs with which model

One of the most common questions we get is whether you can point your favourite editor at any model you like. The pairing rules at RelayVano are explicit, and knowing them up front saves you from debugging a request that never should have failed.

• Claude models work only in the official Claude Code and Claude Desktop clients, version 2.1.285 or newer. They are not wired into other editors, SDKs or command-line tools.
• Codex, Cursor, Cherry Studio, Cline, OpenCode, OpenAI SDKs and cURL pair only with GPT, DeepSeek and the other non-Claude, non-Gemini models.
• Gemini models use the native Gemini protocol only, which means Gemini CLI. Gemini models are not OpenAI-compatible, so an OpenAI-style client will not carry them.

So the flow is short. Install the client that matches your model, paste your RelayVano key into its settings, and point it at the endpoint that belongs to that protocol. If you work across ecosystems, the same key still works; only the client and endpoint combination changes.

How RelayVano meters a request

Every request is broken into parts, and each part is priced on its own. Prompt tokens are the tokens you send. Reply tokens are the tokens the model gives back. Cache stores and cache hits are also priced separately, because writing to a cache and reading from one are not the same operation. On the usage page, every request appears with its charge attached, so you can trace a bill back to a specific call rather than guessing at a monthly total.

The fares below are approximate US dollars per million tokens, with the actual charge made in yuan. For Claude Sonnet 5.5, RelayVano lists $1.25 prompt, $6.25 reply and $0.0889 cache hit, against an official list price of $2, $10 and $0.1 cache hit. For Claude Fable 5.1, the figures are $4.07 prompt, $20.33 reply and $0.2222 cache hit, compared with an official list price of $10, $50 and $0.25 cache hit.

The same pattern holds across the catalogue. GPT-6.1 Sol is $0.1958 prompt, $0.9792 reply and $0.0389 cache hit, while GPT-5.6 Sol is $0.3736 prompt, $2.09 reply and $0.0306 cache hit, with an official list price of $4, $20 and $0.4 cache hit for comparison; that official price is described as an OpenAI promotion stated to run at least until 2026-11-21. Gemini 3.1 Pro is $1.58 prompt, $9.13 reply and $0.1569 cache hit, Gemini 3.8 Flash is $1.20 prompt, $5.40 reply and $0.1194 cache hit, and Gemini 4 Argon is $1.58 prompt, $7.60 reply and $0.0792 cache hit.

Non-Claude, non-Gemini models follow the same structure. DeepSeek V4 Flash is $0.3097 prompt, $1.24 reply and $0.008333 cache hit; Qwen3.8 Max is $0.7431 prompt, $2.23 reply and $0.0889 cache hit; MiniMax M3 is $0.1472 prompt, $0.5889 reply and $0.0292 cache hit. Cache hits are typically the smallest line on a request, which is worth keeping in mind if your prompts repeat a large fixed context.

What happens when a connection drops mid-stream

Streaming responses can end badly. A laptop lid closes, a network flickers, a proxy times out. RelayVano handles that case in a straightforward way: if a caller disconnects mid-stream, the request is still finished and billed on the tokens really used.

This means you are not charged for a reply you never received, and you are not charged for work that never happened. You are charged for what the model actually produced, and that request shows up on the usage page like any other. For teams running long agentic loops, that predictability matters more than a headline rate.

Your prepaid balance never expires either. Credits you add stay available until you spend them, so there is no clock running on your balance and no renewal decision to make before you can keep building.

How top-ups, refunds and referrals work

Top-ups go through Stripe checkout. Cards work everywhere, and WeChat Pay is available where your account supports it. If you change your mind about a top-up, unspent top-ups can be refunded within 9 days.

There is also a referral programme built into the same balance model. You earn 16% of every top-up made by people who sign up through your link, plus 4% one level further down. Earnings can be withdrawn from $5.56 or moved straight into your RelayVano balance, which is handy if you would rather spend them on requests than take them out.

If anything needs clarifying, logged-in members can write to us under Messages in the dashboard. Your key, your balance and your support thread all live in one place, which is the small convenience that makes a relay feel less like plumbing and more like a service.

What a busy day looks like on RelayVano

Here is an example, not a benchmark. Say you send 200 requests a day, and each request carries 1,000 prompt tokens and asks for a 500-token reply, using Claude Sonnet 5.5. That is 200,000 prompt tokens and 100,000 reply tokens per day.

At $1.25 per million prompt tokens, the prompt side comes to $0.25 for the day. At $6.25 per million reply tokens, the reply side comes to $0.625. Together that is about $0.875 a day, or roughly $26.25 across 30 days of the same pattern. At the official list price of $2 and $10 for the same model, the same example would work out to about $0.40 plus $1.00 a day, or $42 across 30 days.

Numbers like these are why the usage page matters. Switch to a lighter model for routine work and a heavier one for the tasks that need it, and the daily figure moves with you. RelayVano does not sell subscriptions or bundled passes; the balance you hold is yours, the tokens you use are metered as they happen, and the cost of a request is visible the moment it completes.

Ready to see it on your own prompts? Create a RelayVano account, claim the $0.40 starter credit, generate one API key, and run a single request through the client that matches your model. The first request is usually the fastest way to understand the whole setup, and the sign-up button below will get you there.