Claude Code on BestRelay API relay: Sonnet 5.5 pay-as-you-go setup

Model portraitsEN2026-10-31

The model you write down is the model that answers

This is an example, and it is set at your own desk on an ordinary weekday. You are using Claude Code with Claude Sonnet 5.5 named in the request, and BestRelay is the independent API relay sitting between the client and the provider. BestRelay forwards each request under the model name the caller wrote, so the model named is the model that answers and nothing is quietly swapped for a cheaper tier. That sentence is the whole commercial argument, and it is worth reading twice before you open the dashboard. The line-up is curated: 20 public models, each listed under its own name with four prices for input, output, cache write and cache read. When a provider retires a model id, BestRelay delists it rather than leaving a stale name in the list for you to trip over. You will find Claude Sonnet 5.5 there, with its own row and its own rates, and you can read them before you spend anything. Nothing here is an official arrangement with Anthropic, OpenAI or Google. BestRelay is an independent relay, and it says so plainly.

Claude Sonnet 5.5 on BestRelay: input $1.39, output $7.01, cache write $1.74, cache read $0.1222 per 1,000,000 tokens.
The official list price for the same four measures is input $3, output $15, cache write $3.75, cache read $0.3.

For example, a day of 500,000 input and 100,000 output tokens costs $1.40 on BestRelay, and 22 such days cost $30.71. With 350,000 of the input read from cache, the day costs $0.95. At the official list price the same day is $3.00, the 22 days $66.00 and the cached day $2.06. The difference on the day is $1.60 and on the 22 days $35.29. These are examples, worked through so you can see the arithmetic rather than take a slogan on trust.

Claude models arrive only through the official Claude clients

Back to the desk. You have two settings to deal with in the client: the address, and a key you create under API keys in the dashboard. The details are in the BestRelay docs, and we will not print either setting here, because a working address pasted into a blog post stops working by lunchtime. What matters is the rule that governs which client may talk to which model. Claude models are served only to the official Claude Code and Claude Desktop clients, version 2.1.285 or newer. They do not work in Cursor, Cherry Studio, Cline, OpenCode, any SDK or cURL. Those tools, like Codex, pair only with GPT, DeepSeek, Kimi, GLM, Qwen and Grok. Gemini models use only their native protocol, through Gemini CLI or direct generateContent calls, and are not OpenAI-compatible.

So if your day involves Claude Code, you are in the supported lane. If it involves one of the other tools, the models beside them are GPT, DeepSeek, Kimi, GLM, Qwen or Grok, and BestRelay lists those too. On one key the limit is 120 requests a minute, and on one account it is 8 calls in flight. Your account can hold several keys, each can be voided at any time, and every call appears in the usage records with its model, tokens and charge. That last point is the one you will thank yourself for at the end of a long session: when a bill arrives, it arrives with the evidence attached, and a key you have finished with can be voided without disturbing the others.

Prepaid credit, charged once and only after the call ends

Money, then. BestRelay runs on a prepaid pay-as-you-go balance. There is no monthly fee, no subscription and no minimum spend. Each call is charged once, after it ends, from the tokens the provider reports, at the rate card prices. Nothing is held in advance. If the provider rejects a request, it is recorded at zero, and when the provider cannot be reached at all, nothing is charged. You are not paying for the privilege of trying something.

A new account opens with $0.35 of credit, which is enough to make a handful of real requests and see the records fill in. For example, that free credit alone covers about 1.6 million GPT-6.1 Sol input tokens, so you can begin with the lighter models and move up once you know what your day actually costs. Top-ups go through Stripe checkout in these amounts: $1.39, $6.94, $13.89, $27.78, $69.44, $138.89. Small ones for a test, larger ones for a stretch of ordinary use.

Unused credit can be refunded on request within 13 days of the top-up, less what was consumed and any payment-processor fees. Gift credit and commission cannot be refunded. That is a plain condition rather than a generous one, and it is easier to accept when the balance starts at $0.35 and grows only when you ask it to.

Referrals, and what the records keep

Some of the desk work is not coding. You invite a colleague, and BestRelay pays you 18% of the top-ups of people who joined through your link, and 4% of the top-ups of the people they invited. Commission can be moved into the balance, or withdrawn to Alipay from $20.83. As an example, on an invited member's top-up of $13.89, the two commissions are $2.50 and $0.5556.

Privacy deserves a paragraph of its own, because API usage tends to accumulate fragments of your working life. The user-written text of requests is kept for 30 days to review abuse and settle billing questions. System prompts and model replies are not stored. So the passage you typed into Claude Code remains on file for a defined window, and the answer itself does not linger. If you want to see how any of this behaves, the usage records give you the model, the tokens and the charge for every call, and BestRelay keeps a curated line-up of 20 public models with four prices each, so there is always a rate card to check a charge against.

Keeping Claude Sonnet 5.5 as the everyday model

Here is the example again, in one paragraph. You sit down, open Claude Code on version 2.1.285 or newer, and point it at BestRelay with a key from the dashboard. You ask for Claude Sonnet 5.5, and Claude Sonnet 5.5 answers, at $1.39 per 1,000,000 input tokens and $7.01 per 1,000,000 output tokens, with cache write at $1.74 and cache read at $0.1222. You work through the day, glance at the usage records, and find the calls listed with their model, tokens and charge. When the provider stumbles, you are not billed for the stumble. When you top up, you top up through Stripe in one of the listed amounts, and you know there is no monthly fee sitting underneath it.

On a heavier afternoon you might step up a tier without changing anything else about your setup, because the curated line-up lets you read the alternatives side by side. Claude Opus 5.5 is listed at input $1.74, output $8.13, cache write $2.01 and cache read $0.0778 per 1,000,000 tokens. Claude Opus 5 sits at input $2.29, output $11.46, cache write $2.85 and cache read $0.1389. Claude Haiku 4.5 is the light one, at input $0.4167 and output $2.01. Each sits under its own name, and each answers under its own name, which is the arrangement that makes a rate card worth reading at all.

That is the whole proposition, and it is deliberately unadorned: name the model, get the model, see the charge, and pay nothing for a failed request. If you would like to try it, an account on BestRelay opens with $0.35 of credit and a curated line-up to read at your own pace.