RelayVano API Relay Pricing in Yuan: Claude Code and Gemini CLI

Field trips EN2026-11-02

Why yuan pricing matters on RelayVano

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. That multilingual setup is the first reason Chinese developers tend to feel comfortable here: on the Chinese pages, prices show in yuan, while the other language pages show approximate dollars. For quick mental arithmetic we use the rough conversion 1 USD ≈ ¥7.2, but the charge itself is always made in yuan. The number you read on the Chinese page is the number that lands on your bill, rather than a figure that drifts with a currency pair.

Clarity is the point. RelayVano sells nothing but a pay-as-you-go balance, and the pricing page lets you read what a prompt token, a reply token and a cache hit cost for each model before you wire anything up. There is no monthly fee to amortise and no seat to buy, which makes RelayVano a sensible fit for side projects, contract work and teams whose usage rises and falls from month to month.

You also get a small runway on day one: new accounts start with $0.40 of credit. That is enough to send a handful of real requests, confirm your client pairing works, and read your first line items on the usage page before deciding how much to top up.

One API key, three protocols, fixed client pairings

A single API key reaches every model through https://api.relayvano.com. Three protocol shapes are available: an OpenAI-compatible /v1 path, the Anthropic /v1/messages path, and the native Gemini /v1beta path. One key, one dashboard, one usage log, and no separate account juggling per model family.

The client pairings are specific, and we would rather state them plainly than leave you to guess:

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 (Gemini CLI); they are not OpenAI-compatible.

So if your daily driver is Claude Code or Claude Desktop, point them at the Anthropic-style path and you are set. If you live in Cursor, Cline, Cherry Studio, OpenCode or Codex, you stay with the OpenAI-compatible path and reach the GPT, DeepSeek, Qwen, Kimi, MiniMax, Hunyuan, Grok and MiMo families there. If you want Gemini, you use Gemini CLI over the native Gemini protocol. Once you know which client you actually use, setup takes minutes rather than an afternoon of troubleshooting.

Because the pairing is fixed, the pricing table stays honest too: a model listed for one protocol does not quietly change behaviour when you paste the same key into a different tool.

What the token fares look like in yuan

All fares below are approximate US dollars per million tokens, shown for comparison with published figures; the charge itself is made in yuan. Prompt tokens, reply tokens and cache hits are priced separately, so a long context and a short answer are never billed as one blended number.

Claude Opus 5.5: $1.65 prompt / $7.82 reply / $0.075 cache hit, against an official list price of $4 / $20 / $0.2 cache hit.
Claude Sonnet 5: $1.27 prompt / $6.36 reply / $0.1111 cache hit, against an official list price of $2 / $10 / $0.2 cache hit.
Claude Fable 5.1: $4.07 prompt / $20.33 reply / $0.2222 cache hit, against an official list price of $10 / $50 / $0.25 cache hit.
GPT-6.1 Sol: $0.1958 prompt / $0.9792 reply / $0.0389 cache hit, against an official list price of $2 / $10 / $0.1 cache hit.
GPT-5.6 Sol: $0.3736 prompt / $2.09 reply / $0.0306 cache hit, against an official list price of $4 / $20 / $0.4 cache hit; that OpenAI promotion is stated to run at least until 2026-11-21.
Gemini 3.1 Pro: $1.58 prompt / $9.13 reply / $0.1569 cache hit. Gemini 3.8 Flash: $1.20 prompt / $5.40 reply / $0.1194 cache hit.
DeepSeek V4 Flash: $0.3097 prompt / $1.24 reply / $0.008333 cache hit. DeepSeek V4 Pro: $1.37 prompt / $4.10 reply / $0.0458 cache hit.
MiniMax M3: $0.1472 prompt / $0.5889 reply / $0.0292 cache hit. Qwen3.7 Plus: $0.1458 prompt / $0.4361 reply / $0.0431 cache hit. MiMo V2.6 Pro: $0.2319 prompt / $0.4653 reply / $0.002778 cache hit.

Two habits make these numbers work harder for you. First, compare a reply token against a prompt token honestly: replies cost several times more, so trimming output length is often the bigger lever than trimming input. Second, look at the cache hit column. On a codebase that you re-ask about all day, a low cache hit rate such as $0.008333 for DeepSeek V4 Flash can change the economics of a whole project, and RelayVano shows that column next to the others instead of hiding it.

Prepaid balance, per-token billing, no monthly fee

RelayVano bills a prepaid balance, and you pay per token. There is no monthly fee, and the balance never expires, so money you top up in a quiet month is still there in a busy one. For a Chinese developer budgeting in yuan, that removes a familiar annoyance: an unused allowance quietly expiring and forcing a repeat purchase.

Every charge is visible. 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 you want to know why Tuesday was more expensive than Monday, you can read the request list instead of guessing from a monthly total.

There is one more behaviour worth knowing about early, because it affects trust in the numbers. If a caller disconnects mid-stream, the request is still finished and billed on the tokens really used. In other words, an interrupted call does not leave work unpaid and does not produce a phantom line item; the meter reflects the tokens the model actually generated on your behalf.

Top-ups in yuan: Stripe, WeChat Pay and refunds

Top-ups go through Stripe checkout. Cards work everywhere, and WeChat Pay is available where your account supports it, so a developer in China can fund the balance with the payment method they already use instead of converting currency through a third-party process. The Chinese pages show the yuan total before you confirm, which keeps the amount you approve and the amount we receive aligned.

If your plans change, unspent top-ups can be refunded within 9 days. That short window is worth planning around: it covers the case where you topped up for a project that got postponed, and it keeps the balance structure honest without a subscription you have to cancel.

For most people the practical rhythm is simple. Start with the $0.40 credit that every new account receives, run a few requests through the client you actually use, read the usage page, then top up an amount you are comfortable holding. Because the balance never expires and there is no monthly fee, you can top up in smaller increments as often as you like.

Starting on RelayVano and bringing colleagues along

Sign-up is the only step left. Create your account, pick your language, copy the API key, and set your client to the base address https://api.relayvano.com with the path that matches your pairing: the OpenAI-compatible /v1 for Codex, Cursor, Cherry Studio, Cline, OpenCode, OpenAI SDKs and cURL, the Anthropic /v1/messages path for Claude Code and Claude Desktop at version 2.1.285 or newer, or the native Gemini /v1beta path for Gemini CLI. Send a test request, watch it appear on the usage page with its charge, and you are operating.

If you would rather not absorb the learning curve alone, RelayVano has a referral programme built for this. People who sign up through your link generate 16% of every top-up for you, and one level further down adds 4%. Earnings can be withdrawn from $5.56 or moved straight into your balance, which is convenient when you would simply rather spend it on tokens.

And when something does not look right, logged-in members write to us under Messages in the dashboard, so support sits in the same place as the usage history and the top-up buttons.

RelayVano is built for developers worldwide, priced in yuan where you are, and simple enough that the hardest part is choosing your model. Create your account, claim the $0.40 starter credit, and take the first request from there.