Claude Desktop on BestRelay: a plain reading of the API relay rates

Coding agents at workEN2026-11-07

BestRelay is an API relay that forwards the model name you wrote

BestRelay is an independent API relay, and the promise is a plain one: each request is forwarded to the model provider under the model name the caller wrote. The model named is the model that answers, and nothing is quietly swapped for a cheaper tier. The catalogue is curated — 20 public models — and each entry carries four prices, for input, output, cache write and cache read. When a provider retires a model id, BestRelay delists it, so the card never points at a name that has gone away.

For people who do not live in a terminal, Claude Desktop is the tidier way in. 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.

Setting a client up takes two things: the address and a key created under API keys in the dashboard. The details are in the docs, and neither of them needs to be printed on a page for readers who would rather click through Claude Desktop than type commands. What remains is the rate card, which is worth reading slowly.

The Claude entries on the BestRelay rate card read line by line

Every model on the BestRelay card is listed under its own name, and the Claude entries are the easiest place to begin. Four numbers per model: what a million input tokens cost, what a million output tokens cost, what a million tokens written to cache cost, and what a million tokens read back from cache cost. Cache read is the small column, and it is the one that rewards anyone who checks the arithmetic before switching a long prompt on.

Claude Fable 5: input $4.17, output $20.97, cache write $5.14, cache read $0.3472; official list price: input $10, output $50, cache read $1.
Claude Fable 5.1: input $4.10, output $20.42, cache write $5.00, cache read $0.2083; official list price: input $10, output $50, cache read $0.25.
Claude Haiku 4.5: input $0.4167, output $2.01, cache write $0.4861, cache read $0.0333; official list price: input $1, output $5, cache read $0.1.
Claude Opus 5: input $2.29, output $11.46, cache write $2.85, cache read $0.1389; official list price: input $5, output $25, cache read $0.5.
Claude Opus 5.5: input $1.74, output $8.13, cache write $2.01, cache read $0.0778; official list price: input $4, output $20, cache read $0.2.
Claude Sonnet 5: input $1.32, output $6.67, cache write $1.67, cache read $0.1167; official list price: input $2, output $10, cache read $0.2.
Claude Sonnet 5.5: input $1.39, output $7.01, cache write $1.74, cache read $0.0944; official list price: input $2, output $10, cache read $0.1.

The rest of the card follows the same grammar. Gemini 2.5 Flash is input $0.2778, output $2.08, cache write $0.2778, cache read $0.0264. Gemini 3.1 Pro is input $1.74, output $9.93, cache write $1.74, cache read $0.1389. DeepSeek V4 Pro is input $1.53, output $4.58, cache read $0.0514. GPT-6.1 Sol is input $0.2083, output $1.11, cache read $0.0431, against an official list price of input $2, output $10, cache read $0.1. BestRelay shows the column, the unit and the name, and leaves the choice with you.

BestRelay bills a prepaid pay-as-you-go balance once per finished call

The money side of BestRelay is deliberately unadorned. You buy credit in advance and spend it as you go: 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 beyond the balance you have already paid for, so an idle account does not quietly drain.

Failure is handled with the same plainness. Requests the provider rejects are recorded at zero, and when the provider cannot be reached nothing is charged. A call that never produced an answer therefore leaves the balance where it stood, which matters when you are paying per token rather than per seat.

Each new account opens with $0.35 of credit, enough to make a small test before you commit anything further. Top-ups run through Stripe checkout, and the amounts on offer are these:

$1.39
$6.94
$13.89
$27.78
$69.44
$138.89

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, which is worth knowing before you treat a referral balance as spendable cash.

Two worked examples show what a month of Claude and GPT traffic costs

Take Claude Sonnet 5.5, as an 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 $0.94. At the official list price, the same day is $2.00, the 22 days $44.00 and the cached day $1.33; the difference on the day is $0.60, and on the 22 days $13.29. The cached figure is the one to notice, because it is where a rate card stops being abstract and starts being a habit.

Now take GPT-6.1 Sol, as an example. A 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; the difference on the day is $1.78, and on the 22 days $39.26. The free credit alone covers about 1.6 million GPT-6.1 Sol input tokens, which is a reasonable way to test the card before topping up.

These are examples, not quotas, and they are the reason BestRelay publishes four prices per model rather than one headline figure. Output is priced separately from input, and cache write is priced separately again. A reader who wants a monthly estimate can build one from the same card the dashboard uses.

Keys, records, referrals and the small print of using BestRelay

An account can hold several keys, each one can be voided at any time, and every call appears in the usage records with its model, tokens and charge. The limits are stated plainly as well: 120 requests a minute on one key, and 8 calls in flight on one account. Both are comfortable numbers for a desktop client and clear enough to plan around.

On privacy, BestRelay keeps the user-written text of requests for 30 days, to review abuse and settle billing questions. System prompts and model replies are not stored. That is the whole retention policy, and it is easier to accept when it is written in one line rather than scattered across pages.

Referrals are also on the card. 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. For people who spend their working day in Claude Desktop rather than a terminal, BestRelay is worth an afternoon of reading the rate card and an account of your own.