BestRelay and Codex: an API relay for weekend coding

BestRelay answers with the model you name

Here is an example of a weekend at your own desk, set out plainly as an example rather than a claim about what your own Saturday will contain. You open a repository before breakfast, you run Codex, and you ask for one file to be tidied. Codex sends that request to a model through an API relay. BestRelay is an independent API relay, and it forwards each request to the model provider under the model name you wrote. The model named is the model that answers, and nothing is quietly swapped for a cheaper tier on the way through.

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. You read the row, you pick the row, and the row is what serves the request. When a provider retires a model id, BestRelay delists it, rather than leaving a name in the list that no longer answers.

The line-up as it stands:

• GPT models: GPT-5.6 Sol, GPT-6 Astra, GPT-6.1 Sol.
• Claude models: Claude Fable 5, Claude Fable 5.1, Claude Haiku 4.5, Claude Opus 5, Claude Opus 5.5, Claude Sonnet 5, Claude Sonnet 5.5.
• Gemini models: Gemini 2.5 Flash, Gemini 3.1 Pro, Gemini 3.8 Flash, Gemini 4 Argon.
• Others: DeepSeek V4 Pro, Kimi K3, Qwen3.8 Max, Grok 4.6, Grok 4.7, GLM-5.3.

For a hobbyist who codes at weekends, that arrangement removes a small but recurring doubt. You are not left wondering which tier your request landed on, or which quieter cousin answered in place of the one you chose. You ask a named model to do a piece of work, and BestRelay forwards the request as written.

BestRelay charges once, after the call ends

Money at BestRelay moves in a plain way. You keep a prepaid pay-as-you-go balance. There is no monthly fee, no subscription and no minimum spend, and nothing is held in advance. When a call ends, that call is charged once, using the token counts the provider reports and the rate card prices for the model you named. A long session does not turn into a monthly bill, and an abandoned afternoon does not turn into a commitment.

The awkward cases are handled with some restraint. A request the provider rejects is recorded at zero. When the provider cannot be reached, nothing is charged. A failed attempt therefore does not quietly eat into the balance you built for the day, which matters when the work is a hobby and the budget is a hobby budget.

A new account opens with $0.35 of credit, enough to see the shape of a working session before you add anything. Top-ups run through Stripe checkout in fixed amounts: $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89. You can start small, and for a weekend coder a small top-up is a sensible way to keep the arrangement tidy.

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. The window is short and the arithmetic is stated, which is about as much as a hobbyist should have to remember about money, and BestRelay keeps the balance visible while you work.

The rate card shows four prices for every model

Each model on the BestRelay list carries four prices per 1,000,000 tokens: input, output, cache write and cache read. Some rows also show the official list price beside them, which makes a comparison easy to check. Claude Sonnet 5.5, for example, is listed at $1.39 input, $7.01 output, $1.74 cache write and $0.0944 cache read, against an official list price of $2 input, $10 output and $0.1 cache read.

The worked figures are worth reading as examples. With Claude Sonnet 5.5, a day of 500,000 input and 100,000 output tokens costs $1.40 on BestRelay, and 22 such days cost $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.

With GPT-6.1 Sol, the same example day of 500,000 input and 100,000 output tokens costs $0.22 on BestRelay, and 22 such days cost $4.74; with 350,000 of the input read from cache the day costs $0.16. At the official list price the same 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. The free $0.35 credit alone covers about 1.6 million GPT-6.1 Sol input tokens.

None of this asks you to remember a plan. You top up when the balance runs low, you read the usage records, and the BestRelay rate card lists four prices for each model on the line-up, so the cost of a weekend session can be read before it begins.

Codex, Claude Code and Gemini CLI each have their own rules

BestRelay serves different clients with different model families, and the boundaries are worth stating early. Codex, like Cursor, Cherry Studio, Cline and OpenCode, pairs with GPT, DeepSeek, Kimi, GLM, Qwen and Grok models. If your weekend habit is to open Codex and point it at GPT-6.1 Sol, that pairing is supported, and the request travels under that model name.

Claude models are served 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, so if you keep a Claude habit, keep it in the client that speaks to it. Gemini models use their native protocol, through Gemini CLI or direct generateContent calls, and are not OpenAI-compatible, which is worth knowing before you spend an afternoon on a connection that will not carry traffic.

Connecting a client takes two settings: the address and a key created under API keys in the dashboard. The details are in the BestRelay docs; the address itself is not something to print in an article. An account can hold several keys, so you can keep one for the Saturday project and void another when it has served its purpose.

BestRelay is an independent API relay. It sits between your client and the model provider, forwarding requests under the name you wrote and recording what the provider reported back, so the arrangement stays legible to the person paying for it.

Limits, records and a 30-day window on request text

BestRelay applies a rate limit of 120 requests a minute on one key, and allows 8 calls in flight on one account. The limits are specific numbers rather than a vague assurance, so you can plan a long session around them.

Each call appears in the usage records with its model, tokens and charge, which is how you check a figure rather than take it on trust. Keys can be voided at any time, and several can live under one account. On privacy, 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.

There is a referral scheme as well. 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. To take one example, on an invited member's top-up of $13.89 the two commissions are $2.50 and $0.5556.

Taken together, BestRelay offers a narrow arrangement a weekend coder can check: the model you name is the model that answers, the rate card is published, and the records are there when you want to look. If that suits how you work, you can open an account and try BestRelay with the $0.35 of starting credit.