RelayVano API Relay Pricing: GPT-5.6 Sol in Quiet and Busy Months

Two months, one pay-as-you-go balance

Some months are quiet. A maintenance script runs twice a week, a side project sits in a drawer, and your token bill is small enough to ignore. Then a launch lands, a team adds a new integration, and the same account starts sending requests every hour. RelayVano is built for both seasons because there is no subscription to renew and no monthly fee to justify: you top up a prepaid balance, and every request is charged per token as it happens.

The balance never expires, so quiet-month credit is not lost while you wait for the busy month to arrive. Prompt tokens, reply tokens, cache stores and cache hits are priced separately, and every request shows up with its charge on the usage page. That page is where the shape of a month becomes visible: which models you actually called, how many tokens each direction used, and what the cache saved you.

On RelayVano, one API key reaches every model through https://api.relayvano.com: an OpenAI-compatible /v1 path, Anthropic /v1/messages, and the native Gemini /v1beta path. That single key is the practical reason a busy month does not turn into a billing scramble. You keep one account, one balance and one usage history, even when your workflow spans several model families.

What a quiet month looks like on RelayVano

Take a quiet month as 20 requests a day — this is an example, not a measurement of any customer. Assume each request carries 1,000 prompt tokens and asks for a 500-token reply. With GPT-5.6 Sol at $0.3736 prompt and $2.09 reply per million tokens, a day costs about $0.0075 in prompt tokens and $0.0209 in reply tokens, roughly $0.028. Across 30 days that is about $0.85.

Run the same example on MiniMax M3, priced at $0.1472 prompt and $0.5889 reply per million tokens, and the daily figure is about $0.0029 plus $0.0059, so roughly $0.009 a day and about $0.26 for the month. The gap between the two models is not a gap between two billing systems; it is simply the difference in what each model costs to run.

MiMo V2.6 Pro behaves similarly in this example: $0.2319 prompt and $0.4653 reply per million tokens put the day near $0.046 plus $0.047, or about $0.093, and around $2.79 over 30 days. Quiet months stay inexpensive on RelayVano mainly because nothing is charged for capacity you did not use.

What a busy month looks like

Now widen the example. Say you send 200 requests a day — again, purely an example — with the same 1,000 prompt tokens and 500 reply tokens per request. GPT-5.6 Sol works out to about $0.0747 in prompt tokens and $0.209 in reply tokens per day, roughly $0.284. Over 22 working days that lands near $6.24; across a full 30 days, about $8.51.

Switch that busy workload to MiniMax M3 and the same volume costs roughly $0.0294 in prompt tokens plus $0.0589 in reply tokens each day, about $0.088, or around $2.65 over 30 days. On MiMo V2.6 Pro, the same traffic is about $0.0464 plus $0.0465 a day, close to $0.093, and about $2.79 for 30 days.

The point is not that one model suits every workload. It is that a spike has a price you can estimate before it arrives, and a prepaid balance always tells you what is left. If you are weighing RelayVano against a vendor's subscription, per-token pricing means a heavy month costs what you actually send, while a light month costs very little.

Prompt, reply and cache tokens priced separately

Reply tokens are usually the largest line in these examples, which is why the reply rate deserves attention. On RelayVano, prompt tokens, reply tokens, cache stores and cache hits are each priced on their own, and the usage page breaks the charge down per request. Nothing is averaged into a monthly figure you have to interpret later.

Cache hits are where careful habits show up. MiMo V2.6 Pro lists a cache hit at $0.002778 per million tokens, and DeepSeek V4 Flash lists $0.008333. For a workload that repeatedly reuses a fixed context, the cached portion of the prompt can cost a fraction of a fresh prompt read, and you can see that effect directly on the usage page rather than guessing at it.

Claude pricing follows the same structure. Claude Sonnet 5.5 is listed at $1.25 prompt, $6.25 reply and $0.0889 cache hit per million tokens, against an official list price of $2 / $10 / $0.1 cache hit. Claude Fable 5.1 is $4.07 / $20.33 / $0.2222, compared with an official $10 / $50 / $0.25. Claude Haiku 4.5 is $0.3958 / $1.90 / $0.0319, against an official $1 / $5 / $0.1.

Cache stores are charged separately from cache hits, so a workload that writes a large context once and reads it many times is worth checking on the usage page. And if a caller disconnects mid-stream, RelayVano still finishes the request and bills on the tokens really used — a useful detail when a client drops a long generation.

Client pairing that keeps setup predictable

Cost control only works when the request path is predictable. On RelayVano the client pairing is fixed, with no exceptions. Claude models work only in the official Claude Code and Claude Desktop clients, version 2.1.285 or newer. Gemini models use the native Gemini protocol only, in Gemini CLI, and they are not OpenAI-compatible.

Codex, Cursor, Cherry Studio, Cline, OpenCode, the OpenAI SDKs and cURL pair only with GPT, DeepSeek and the other non-Claude, non-Gemini models. That group includes GPT-5.6 Sol, GPT-6 Astra, GPT-6.1 Sol, DeepSeek V4 Flash, DeepSeek V4 Pro, Qwen3.7 Plus, Qwen3.8 Max, Hunyuan 4, Grok 4.6, Grok 4.7, Kimi K3 and MiMo V2.6 Pro.

Setup is therefore short: create a key, point your client at the right path, then check the first request on the usage page. Because one key serves all three paths, moving a task from a GPT model to a DeepSeek model — or from a Claude model to a Gemini model — does not mean managing another account. It means adjusting one line of configuration and watching the per-token charge.

Prepaid balance, six languages, and your first month

RelayVano sells prepaid balance, not a subscription. Top-ups run through Stripe checkout, so cards work everywhere, plus WeChat Pay where your account supports it. If you overfill the balance, unspent top-ups can be refunded within 9 days. New accounts start with $0.40 of credit, enough to send a few real requests and read the usage page before you commit to a larger top-up.

Referrals work in the same transparent way: you earn 16% of every top-up 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 usage.

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. If you need help, logged-in members write to us under Messages in the dashboard.

Quiet months and busy months both fit inside a pay-as-you-go budget. Start with the $0.40 credit, send a few requests, read the usage page, and then decide how you want the next month to look. Sign up for RelayVano and put your own numbers on the page.