RelayVano API Relay Pricing for Freelancers Using GPT-5.6 Sol

One key, many models, and a site in six languages

If you bill clients per project, the last thing you want is a complicated AI bill to explain at the end of the month. You want to know what you spent, which project it belongs to, and what to top up next. RelayVano was built for that way of working: it is an independent relay made for developers worldwide, and you pay for the tokens you actually use.

RelayVano comes in six languages across the whole site, dashboard and docs: 中文, English, Español, हिन्दी, Português and Français. Prices appear in yuan on the Chinese pages and in approximate dollars on the others, so a developer in any of those markets can read the numbers in the currency they already think in. For a freelancer working with international clients, that small detail removes a layer of guesswork before you even start.

The core idea is simple. One API key reaches every model through https://api.relayvano.com, using the OpenAI-compatible /v1 path, the Anthropic /v1/messages path, or the native Gemini /v1beta path. You are not juggling a separate vendor account, a separate card and a separate invoice for each model family you want to try. You create one account, generate one key per client project, and point your tools at the same base address.

Matching your client to the right model

Here is the part worth reading carefully, because each tool speaks its own protocol and RelayVano keeps the pairing exact rather than approximate. Claude models work only in the official Claude Code and Claude Desktop clients, and you need version 2.1.285 or newer. Codex, Cursor, Cherry Studio, Cline, OpenCode, the 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, and they are not OpenAI-compatible.

For a freelancer, this reads as a short decision rather than a research project. If the client asked for a Claude-backed workflow, open Claude Code or Claude Desktop, update it if needed, and paste your RelayVano key into the Anthropic-compatible configuration. If the deliverable is a code review or an automation built on GPT, DeepSeek, Qwen, MiniMax or another non-Claude, non-Gemini model, the same key fits the OpenAI-compatible route in Codex, Cursor, Cherry Studio, Cline, OpenCode, the OpenAI SDKs or cURL.

Setup itself is a few sentences of work. Create your account, open the dashboard and generate a key for the project you are billing. Set the base address to https://api.relayvano.com, choose the path that matches your client, and paste the key. Because the key is per project, you can hand a client an itemised cost for their own work instead of a blended number that covers everything you did that month.

What per-key usage looks like on the dashboard

RelayVano sells pay-as-you-go balance and nothing else. There is no monthly fee, the balance never expires, and you are charged per token. New accounts start with $0.40 of credit, which is enough to make a first request and see exactly how the numbers appear before you commit to a larger top-up.

The usage page is where per-project billing becomes concrete. Prompt tokens, reply tokens, cache stores and cache hits are priced separately, and every request shows up with its charge on the usage page. When a client asks where the money went, you can point at the request, not at a vague estimate. That is the difference between a quote you can defend and one you have to apologise for.

There is also a practical safeguard for long-running jobs. If a caller disconnects mid-stream, the request is still finished and billed on the tokens really used. In other words, a dropped connection does not leave you paying for work that never happened, and it does not leave you wondering whether the provider quietly billed a partial run either. What was genuinely consumed is what gets charged.

Working out what a busy month would cost

To make the pricing tangible, consider an example. Say you send 200 requests a day, each with 1,000 prompt tokens and 500 reply tokens. That is an example with stated assumptions, not a measurement of your own work. It works out to 200,000 prompt tokens and 100,000 reply tokens per day, or 6 million prompt tokens and 3 million reply tokens over 30 days.

On GPT-6.1 Sol, at $0.1958 prompt and $0.9792 reply per million tokens, that example comes to roughly $1.17 for prompt tokens and $2.94 for reply tokens over the month, about $4.11 in total. At GPT-5.6 Sol, priced at $0.3736 prompt and $2.09 reply, the same example lands near $2.24 plus $6.27, or about $8.51. The point is not that one model is right for everything, but that you can put a defensible number next to a proposal before you start.

For lighter tasks the arithmetic gets friendlier. MiniMax M3 is listed at $0.1472 prompt and $0.5889 reply per million tokens, and Qwen3.7 Plus at $0.1458 prompt and $0.4361 reply. DeepSeek V4 Flash sits at $0.3097 prompt and $1.24 reply, while Qwen3.8 Max is $0.7431 prompt and $2.23 reply. Where a client wants a premium Claude model, Claude Opus 5.5 is $1.65 prompt and $7.82 reply, and Claude Fable 5.1 is $4.07 prompt and $20.33 reply, compared with an official list price of $10 and $50 for that same prompt and reply pair.

Cache hits matter more than most newcomers expect, especially on repeated context. Claude Fable 5.1 lists a $0.3125 cache hit rate, Gemini 3.1 Pro lists $0.1569, and MiMo V2.6 Pro lists $0.002778. Since cache stores and cache hits are billed on their own lines, a workflow that reuses a large fixed context can cost noticeably less than one that resends it from scratch. Your usage page shows which of these lines each request actually hit.

Top-ups, refunds and earning from referrals

Adding money is deliberately low-friction. Top-ups go through Stripe checkout, so cards work everywhere, plus WeChat Pay where your account supports it. If a client pays you up front and then cancels the scope, unspent top-ups can be refunded within 9 days, which makes a prepaid balance far easier to offer as a client-facing arrangement.

RelayVano also rewards people who bring colleagues along. If someone signs up through your link and tops up, you receive 16% of every top-up, and 4% one level further. Earnings can be withdrawn from $5.56 or moved into your balance. For a freelancer who talks to other developers every day, that turns ordinary word-of-mouth into a small recurring income stream rather than a favour you do for free.

Support does not leave you guessing either. Logged-in members write to us under Messages in the dashboard, so a question about a charge, a model pairing or a client invoice stays attached to your account history. You do not have to describe your whole setup from scratch to get an answer.

Signing up for your next billed project

RelayVano suits freelancers who bill per project because it keeps the mechanics plain: one account, one key per project, pay per token, no monthly fee, and a usage page that itemises every request. Whether your client wants GPT-5.6 Sol, Qwen3.7 Plus, MiniMax M3, DeepSeek V4 Flash, a Claude model in Claude Code or Claude Desktop, or Gemini 3.8 Flash in Gemini CLI, the same relay handles the route and the accounting.

Start with the $0.40 of new-account credit, send a test request, and watch the charge appear on the usage page. That first small step is usually the moment the idea becomes obvious. When you are ready, create your account and set up a key for the next project you are quoting.