BestRelay and GPT-6.1 Sol: an API relay with clear pricing

BestRelay forwards each request to the model you name

Consider a sample Tuesday at your own desk, treated throughout as an example. You have a task in mind and a model in mind with it: Claude Opus 5.5, say, or GPT-6.1 Sol. With BestRelay, each request is forwarded to the model provider 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 on the way through. BestRelay is an independent API relay, and its line-up is curated: 20 public models, each listed under its own name with four prices, covering input, output, cache write and cache read. When a provider retires a model id, BestRelay delists it rather than leaving a stale entry in the table for you to trip over later.

Billing is deliberately plain. You open a prepaid pay-as-you-go balance, and 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 for the model you selected. Nothing is held in advance, so an unused balance stays unused rather than being committed to a plan. If the provider rejects a request, it is recorded at zero. If the provider cannot be reached, nothing is charged. Every new account opens with $0.35 of credit, which is enough to make a few real calls before you top up.

On the sample Tuesday, the morning draft, the afternoon edit and the evening summaries all follow the same rule: the model on the label is the model that replies, and the dashboard shows what each call cost.

At the per-token sum, the official list price asks for more

Take the sample day as an example: you write 500,000 input tokens and receive 100,000 output tokens. On Claude Sonnet 5.5, BestRelay charges $1.40 for that day, and 22 such days come to $30.71. If 350,000 of the input is read from cache, the day costs $0.94. At the official list price, the same day is $2.00, the 22 days $44.00 and the cached day $1.33, so the difference is $0.60 on the day and $13.29 across the 22 days. The rate card behind it lists Claude Sonnet 5.5 at $1.39 input and $7.01 output, against an official list price of $2 and $10.

The gap widens on some entries, and the arithmetic stays the same. As an example, GPT-6.1 Sol costs $0.22 for the same day on BestRelay and $4.74 for 22 such days, with a cached day at $0.16. At the official list price that day is $2.00, the 22 days $44.00 and the cached day $1.33, a difference of $1.78 on the day and $39.26 over the 22 days. GPT-6.1 Sol is listed at $0.2083 input and $1.11 output, with an official list price of $2 and $10. The $0.35 of free credit alone covers about 1.6 million GPT-6.1 Sol input tokens.

Nothing here sits behind a monthly commitment. You pay for tokens, on the model you asked for, and the usage records keep the arithmetic visible line by line rather than summarised into a monthly invoice you have to take on trust.

Connecting takes two settings, and the keys stay under your control

In this example, the desk work starts with configuration rather than paperwork. A client needs two settings: the address and a key. You create the key under API keys in the dashboard, and the details are in the docs. That is the whole setup, and the rest of the day is ordinary work: the client sends a request, BestRelay forwards it under the model name you wrote, and the reply comes back from that model.

An account can hold several keys, which matters on a sample day when one key is used for drafting and another for a script that runs unattended. Any key can be voided at any time, so a key that wanders into a log or a repository can be retired in the afternoon without disturbing the others. Every call appears in the usage records with its model, its tokens and its charge, which means the day's total can be checked against the rate card rather than guessed at from memory.

The limits are stated plainly: 120 requests a minute on one key, and 8 calls in flight on one account. On the sample desk that is a workable boundary for one person, and it gives an unattended script somewhere to stop.

Privacy is handled with restraint. The user-written text of requests is kept for 30 days so abuse can be reviewed and billing questions settled. System prompts and model replies are not stored.

Client compatibility is narrow on purpose, so the pairing is never wrong

The sample day has a client question in it, and BestRelay answers it by protocol rather than by guesswork. 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. That is a limitation, plainly stated, and it means a Claude model is never offered to a tool that cannot speak to it.

The rest of the line-up works differently. Codex and tools like it 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. In the example, that is a routing decision made once at the start of the week, rather than an argument with an error message later on.

Because the line-up is curated to 20 public models, each listed under its own name with four prices, the choice stays readable. Claude Opus 5.5 sits at $1.74 input and $8.13 output. Gemini 3.1 Pro sits at $1.74 input and $9.93 output. Grok 4.7 sits at $0.625 input and $1.94 output. The model you name is the model that answers, and the client you use decides which names are in reach.

Top-ups, refunds and referrals keep the balance ordinary

Money in, money out, and the middle of it is small. Top-ups run through Stripe checkout in fixed amounts: $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89. There is no monthly fee, no subscription and no minimum spend, so the size of a top-up reflects how much you expect to use rather than a tier you have been placed on.

Refunds are available on request within 13 days of a top-up, less what was consumed and any payment-processor fees. Gift credit and commission cannot be refunded, which is worth knowing before you treat a referral balance as spendable in the same way as a top-up.

Referrals pay their own small way. You receive 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.

Back at the sample desk, the day ends with a balance that can be read line by line: model, tokens, charge, and a top-up receipt if you added one. If you would like to see how that arithmetic looks on your own work, BestRelay is ready when you are.