BestRelay with Claude Code: an API relay for burst coding

BestRelay forwards each request to the model you named

Here is an example, written to you, of a coding day at your own desk. You do not sit down for one fixed stretch; you work in bursts, and the shape of the day belongs to you.

In the morning you open BestRelay and ask Claude Sonnet 5.5 about a function that will not behave. BestRelay forwards the request to the model provider under the model name you wrote, so the model named is the model that answers, and nothing is quietly swapped for a cheaper tier before your question leaves the desk.

Around midday you switch to Claude Opus 5.5 for a harder refactor, and in the afternoon you try Claude Haiku 4.5 for a small edit you keep making. Each of those choices is yours to make by name, and each one is priced in the same published place.

The point of this example is not that the day resembles any other day, but that the accounting stays plain while the day does not. BestRelay lists 20 public models in a curated line-up, and each one is listed under its own name with four prices per 1,000,000 tokens: input, output, cache write and cache read.

You never have to hold a subscription in your head, because there is no monthly fee, no subscription and no minimum spend. You keep a prepaid balance, and each call is charged once, after it ends, from the tokens the provider reports.

When a provider rejects a request, it is recorded at zero. When the provider cannot be reached, nothing is charged at all. A burst that goes wrong does not quietly turn into a bill.

Twenty curated models, four prices each, nothing hidden

BestRelay's line-up is curated rather than sprawling: 20 public models, each shown under its own name with four prices per 1,000,000 tokens — input, output, cache write and cache read.

That structure makes comparison quick. Claude Haiku 4.5 is listed at $0.4167 input and $2.01 output. GPT-6.1 Sol is listed at $0.2083 input and $1.11 output. Gemini 2.5 Flash is listed at $0.2778 input and $2.08 output. Claude Opus 5.5 sits at $1.74 input and $8.13 output, while Claude Sonnet 5.5 is listed at $1.39 input and $7.01 output.

Cache pricing is published in the same row rather than tucked into a footnote. Claude Sonnet 5.5 lists cache write at $1.74 and cache read at $0.0944; Claude Fable 5.1 lists cache read at $0.2083; Gemini 3.1 Pro lists cache write at $1.74 and cache read at $0.1389.

The line-up also includes Claude Fable 5, Claude Opus 5, Claude Sonnet 5, DeepSeek V4 Pro, Kimi K3, Qwen3.8 Max, Grok 4.6, Grok 4.7, GLM-5.3 and Gemini 3.8 Flash, each priced in the same manner.

Curation has a second consequence. When a provider retires a model id, BestRelay delists it, so the names on the pricing page reflect what can still be called rather than a historical catalogue kept for appearance's sake.

For a coder who works in bursts, this matters less about shelf depth than about certainty: the name you type is the name that answers, and the four numbers beside it are the numbers you will pay. BestRelay keeps the two facts in one row so you can check a model before you spend a token on it.

A prepaid balance, charged once per finished call

BestRelay bills on a prepaid pay-as-you-go balance. There is no monthly fee, no subscription and no minimum spend, and nothing is held in advance: each call is charged once, after it ends, from the tokens the provider reports, at the rate card prices.

Every new account opens with $0.35 of credit, which is enough to make a small number of calls before you add to the balance and get on with the day.

Top-ups go through Stripe checkout in fixed amounts: $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89. You choose the amount that suits the size of your bursts, from a single small top-up to a larger balance for a busy week.

Failed calls are handled without drama. A request the provider rejects is recorded at zero, and if the provider cannot be reached, nothing is charged. You are not asked to pay for a call that never produced tokens.

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.

BestRelay also carries a referral scheme: 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.

For example, on an invited member's top-up of $13.89, the two commissions are $2.50 and $0.5556.

For someone who codes in bursts, this shape is comfortable. You are not committing to a month of anything; you are holding a small balance and drawing it down as the work arrives.

Claude Code and Claude Desktop are the route for Claude models

Claude models on BestRelay are served only through Claude Code and Claude Desktop, version 2.1.285 or newer. That is a narrow door, and it is deliberate.

Gemini models use only their native protocol, through Gemini CLI or direct generateContent calls, and are not OpenAI-compatible. Codex and comparable tools pair only with GPT, DeepSeek, Kimi, GLM, Qwen and Grok models.

Connecting a client takes two settings: the address, and a key you create under API keys in the dashboard. The details are in the docs. We do not print the address or any setting here, and you would be wise to keep your key to yourself.

The limits are stated plainly too: 120 requests a minute on one key, and 8 calls in flight on one account. For desk-sized bursts that is room to work, and it is a ceiling you can see rather than guess at.

An account can hold several keys, and any of them can be voided at any time. Every call appears in the usage records with its model, its tokens and its charge, so after a busy afternoon you can read back what each burst cost and which model produced it.

On privacy, the position is specific: the user-written text of requests is kept for 30 days to review abuse and settle billing questions, while system prompts and model replies are not stored.

BestRelay is an independent API relay, and that independence shapes the whole offer: it forwards requests, keeps the books, and publishes the prices it charges.

What a burst costs, shown as an example

Here is the arithmetic of an example day, using 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. With 350,000 of the input 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 same example with GPT-6.1 Sol comes out at $0.22 for the day and $4.74 for 22 days, or $0.16 for the cached day, against $2.00, $44.00 and $1.33 at the official list price — a difference of $1.78 on the day and $39.26 across the 22 days.

The free credit alone covers about 1.6 million GPT-6.1 Sol input tokens, which is a comfortable place to start before you add to the balance.

None of this requires a plan, a promise or a minimum. You name a model, the named model answers, and the tokens are counted once and charged once.

If you like the shape of this example — a small balance, a clear line-up, and bills that follow the tokens rather than a subscription — BestRelay is worth opening an account on and trying for a day of your own.