RelayVano API Relay: One Balance for GPT, Claude and Gemini

Fare watch EN2026-10-31

One RelayVano balance, many models

Keeping track of separate accounts can make AI development feel heavier than it needs to be. You may have one login for a GPT workflow, another for Claude, and a third for Gemini, each with its own billing page, key format and top-up process. RelayVano gives you a different starting point: one prepaid balance, one API key and one usage page for the models you actually use.

RelayVano is an independent relay made for developers worldwide. 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 the interface can feel familiar whether you are setting up a project in Beijing, Madrid, Mumbai, São Paulo, Paris or somewhere else.

Our aim is straightforward: give developers a practical way to reach GPT, Claude, Gemini, DeepSeek, Qwen, Kimi, MiniMax, Grok, Hunyuan and MiMo models without managing a separate vendor account for every one. You add credit to your RelayVano balance, then spend it as requests are made. There is no monthly fee, and the balance never expires.

What the API key can and cannot reach

All model traffic goes through our base address, https://api.relayvano.com. From there, the request format depends on the model family. GPT, DeepSeek and the other non-Claude, non-Gemini models use the OpenAI-compatible /v1 route. Claude models use Anthropic /v1/messages, and Gemini models use the native Gemini /v1beta protocol.

Those formats matter because they decide which tools you can use with each model. Claude models work only in the official Claude Code and Claude Desktop clients, version 2.1.285 or newer. 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, in Gemini CLI; they are not OpenAI-compatible.

For Claude Code or Claude Desktop, point the client at the RelayVano base address and use the Anthropic /v1/messages route with your RelayVano key.
For Codex, Cursor, Cherry Studio, Cline, OpenCode, OpenAI SDKs or cURL, use the OpenAI-compatible /v1 route with GPT, DeepSeek or another non-Claude, non-Gemini model.
For Gemini CLI, use the native Gemini /v1beta route. Do not expect Gemini to behave like an OpenAI-compatible model.

There is no need to guess whether a particular combination is supported. The pairing is specific, and the same key can serve each supported route. That is the convenience of a shared balance: the billing arrangement stays the same even when the model or client changes.

Pay per token, with every charge visible

RelayVano uses a prepaid balance and pay-per-token billing. Prompt tokens, reply tokens, cache stores and cache hits are priced separately, and every request appears with its charge on the usage page. Instead of searching for an invoice at the end of a period, you can see what each call cost while you are building.

Consider an example: suppose you send 200 requests a day to MiniMax M3. Using the published fare of $0.1472 per million prompt tokens, $0.5889 per million reply tokens and $0.0292 per million cache-hit tokens, the daily cost depends on how many tokens those requests contain and how many are cache hits. If each example request used 1,000 prompt tokens and 1,000 reply tokens, the daily amount would be 200,000 prompt tokens at $0.1472 per million, plus 200,000 reply tokens at $0.5889 per million, which comes to about $0.1472 of prompt charges and $0.11778 of reply charges, or roughly $0.26498 per day. This is only an example, not a typical workload, and the usage page remains the source of truth for your own traffic.

The same per-token approach applies across the catalogue. GPT-6.1 Sol is listed at $0.1958 prompt and $0.9792 reply, while Claude Sonnet 5.5 is $1.25 prompt and $6.25 reply. Gemini 3.8 Flash is $1.20 prompt and $5.40 reply. You can compare these fares with the official list prices quoted for models such as GPT-6.1 Sol, where the official figure is $2 prompt and $10 reply per million tokens.

Cache hits can also make a meaningful difference in longer-running sessions. Claude Haiku 4.5 is listed at $0.3958 for prompt tokens, $1.90 for reply tokens and $0.0319 for cache hits, against official list prices of $1 and $5. That is another reason to check the usage page: you can see which part of a request is being charged and where caching is helping.

Top-ups, refunds and a small head start

Adding credit is handled through Stripe checkout. Cards work everywhere, plus WeChat Pay where your account supports it. Unspent top-ups can be refunded within 9 days, which gives you room to try the service and check the usage figures before committing more of your budget.

New accounts start with $0.40 of credit. That is enough to send a few small requests, confirm that your chosen client reaches the intended route, and inspect the first entries on the usage page before you top up.

There is no subscription to maintain and no monthly fee to justify. If a project pauses, the balance does not expire. If you need help, logged-in members can write to us under Messages in the dashboard, where the conversation stays connected to your account.

Referrals that share part of the top-up

If you prefer to try RelayVano before spending your own credit, you can invite other developers through your referral link. You receive 16% of every top-up made by people who sign up through your link, and 4% one level further. Earnings can be withdrawn from $5.56 or moved into the balance to cover your own requests.

For example, if someone who signs up through your link adds $20, your referral share is $3.20. That can be moved straight into the balance for API usage or withdrawn once you reach the $5.56 threshold. The example uses the published referral rate; actual earnings follow each eligible top-up recorded through your link.

Start with one key and check the usage page

RelayVano is built to make the first steps easy. Create an account, use the $0.40 starter credit, and connect the client that matches your model. Send a small request, then open the usage page to see the charge recorded. From there, you can decide which model fits the job and whether the per-token fares work for your volume.

You can also plan for the edge case that often worries teams: if a caller disconnects mid-stream, the request is still finished and billed on the tokens really used. That keeps the accounting consistent with the work already performed.

One balance, one key and a usage page that shows the numbers are a simple arrangement for developers who would rather spend their time on the application than on account administration. Sign up for RelayVano, try the starter credit, and see how far your next project can go.