RelayVano API Relay Pricing: GPT-6.1 Sol vs Official List Price

What the GPT-6.1 Sol price comparison actually shows

Most developers meet a relay with one question in mind: what will this model cost me at the end of the month? We built RelayVano so that the answer is visible before you commit to anything, not buried in a sales conversation. RelayVano is an independent relay made for developers worldwide, and every fare is listed per million tokens with prompt, reply and cache-hit pricing shown side by side.

For GPT-6.1 Sol, our pay-as-you-go fares are $0.1958 per million prompt tokens, $0.9792 per million reply tokens and $0.0389 per million cache-hit tokens. The official list price we quote for comparison is $2 / $10 / $0.1 cache hit. Put plainly, the prompt and reply figures sit roughly 90% below the official list price, while the cache-hit figure is about 61% below it.

• GPT-6.1 Sol prompt tokens: $0.1958 per million, against an official list price of $2.
• GPT-6.1 Sol reply tokens: $0.9792 per million, against an official list price of $10.
• GPT-6.1 Sol cache-hit tokens: $0.0389 per million, against an official list price of $0.1.

These are approximate US dollar figures for planning purposes (1 USD ≈ ¥7.2), and the actual charge is made in yuan. That distinction matters if you are comparing a subscription from a vendor with a pay-as-you-go balance: one is a recurring commitment, the other is usage you pay for as it happens.

It also matters that a headline price is rarely the whole bill. Long system prompts, large retrieved documents and repeated context all change the mix of tokens you send and receive. RelayVano prices prompt tokens, reply tokens, cache stores and cache hits separately, so you can read the usage page and see which part of your traffic is actually driving the charge.

Common assumptions about relay billing, and how we handle it

A common assumption is that a relay requires a subscription, a seat fee or a minimum monthly commitment. RelayVano does not work that way. There is no monthly fee. You add credit to a prepaid balance, and you are charged per token as requests arrive. The balance never expires, so an unused remainder is not lost when a billing cycle ends.

Another assumption is that a small credit balance is a trap: you top up, experiment, and then discover that you must top up again to keep working. New accounts start with $0.40 of credit, which is enough to send a first request and confirm that your client, protocol and model pairing behave the way you expect before you commit to a larger top-up.

The third assumption is about visibility. On RelayVano, every request shows up with its charge on the usage page. You can review what you sent, what came back and what it cost. If a caller disconnects mid-stream, the request is still finished and billed on the tokens really used. We would rather show you the real consumption than quietly drop an interrupted call and leave you guessing whether the money was spent.

For a team that is still evaluating, this makes the decision concrete. You can start with the starter credit, watch a handful of requests appear on the usage page with their charges, and only then decide how much to add. There is no lock-in period to wait out and no cancellation ritual to remember.

How one API key fits the clients you already use

One API key reaches every model through https://api.relayvano.com. Behind that single base address sit three protocol styles: an OpenAI-compatible /v1 endpoint, the Anthropic /v1/messages endpoint and the native Gemini /v1beta endpoint. You point your client at the base address, choose the model and start working.

What we do not do is blur the boundaries between protocols, because that is where assumptions usually go wrong. GPT-6.1 Sol, like our other GPT, DeepSeek and non-Claude, non-Gemini models, pairs with Codex, Cursor, Cherry Studio, Cline, OpenCode, OpenAI SDKs and cURL. If you are coming from one of those tools, the setup path is short: set the base address, add your key and select the model.

• 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. They are not OpenAI-compatible.
• GPT, DeepSeek and other non-Claude, non-Gemini models pair with Codex, Cursor, Cherry Studio, Cline, OpenCode, OpenAI SDKs and cURL.

Keeping these pairings exact is deliberate. A relay that advertises one key for everything but quietly changes the protocol underneath can cost you an afternoon of debugging. RelayVano keeps each model on the route it expects, so the setup you read about is the setup you get.

A six-language relay for developers working worldwide

RelayVano is an independent relay made for developers worldwide, and the language question came up early. Documentation written for one audience tends to assume a context that does not travel well, so the whole site, dashboard and docs come in six languages: 中文, English, Español, हिन्दी, Português, Français.

Pricing follows the same practical logic. Prices show in yuan on the Chinese pages and in approximate dollars on the others, so you can read the numbers in the currency that matches how you actually pay. Nothing about the underlying account changes with the interface language: the same key, the same usage page and the same prepaid balance are there in each version.

For teams spread across several time zones, this matters less about translation than about consistency. A colleague in Spain, a teammate in Brazil and a developer in India can each work in a familiar interface and still land on the same model fares. Support is handled in the same place too: logged-in members write to us under Messages in the dashboard, so the conversation stays attached to the account and the usage history rather than scattered across inboxes.

Because RelayVano is independent, we are not tied to one vendor's roadmap or one region's commercial model. That independence is what lets us publish a straightforward per-token fare list and keep it in front of you, updated as the lineup changes.

Top-ups, refunds and referrals without fine print

Adding credit is deliberately unremarkable. Top-ups go through Stripe checkout: cards everywhere, plus WeChat Pay where your account supports it. You choose the amount, complete the checkout and the balance is ready for the next request. There is no seat count to negotiate and no term to agree to.

If you add more than you need, unspent top-ups can be refunded within 9 days. We would rather you feel comfortable testing the service than treat a top-up as a one-way door, especially when you are still comparing a pay-as-you-go relay against a subscription you already pay for.

There is also a way to earn from usage you send us. Referrals pay 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, so you can either take them out or spend them on your own traffic. A developer who is already running requests through RelayVano can put that habit to work without turning into a full-time sales team.

None of this requires a long contract. It is a balance, a key and a fare list. That is the whole commercial model, and it is why the GPT-6.1 Sol comparison above is so easy to check for yourself.

What to expect after you sign up

Here is the shape of a typical first session. You create an account and find $0.40 waiting as starter credit. You pick the client that matches your workflow, set the base address to https://api.relayvano.com, add your key and select GPT-6.1 Sol. You send a request, and the charge appears on the usage page with the prompt, reply and cache components broken out.

From there, the decision is yours: keep the balance topped up as your traffic grows, or move to a larger top-up with the refund window still open if you overestimated. If you later want to compare across the lineup, RelayVano also carries DeepSeek, Gemini, Kimi, MiniMax, Qwen, Hunyuan, Grok and MiMo models, each priced per token with the same separate prompt, reply, cache-store and cache-hit breakdown.

We think the practical proof is straightforward. Put GPT-6.1 Sol next to its official list price, look at $0.1958 per million prompt tokens and $0.9792 per million reply tokens, then check the usage page after your first real workload. If the numbers look right to you, the next step is small: sign up, use the $0.40 of credit, and see how your own traffic adds up.