Gemini CLI on BestRelay: pay-as-you-go API relay setup

BestRelay forwards Gemini CLI requests to the model you named

BestRelay is an independent API relay, and its central claim is narrow enough to check in a single line of a request log: the model you name is the model that answers. Each request is forwarded to the model provider under the model name the caller wrote, so nothing is quietly swapped for a cheaper tier behind the scenes. The line-up is curated rather than sprawling — 20 public models, each listed under its own name with four prices attached: input, output, cache write and cache read.

If you use Gemini CLI and ask for Gemini 3.1 Pro, BestRelay forwards the request under that name and records it under that name. When a provider retires a model id, BestRelay delists it, which keeps the catalogue honest: the entries you can see are entries you can call. The same rule applies across the line-up, which spans Claude Fable 5, Claude Opus 5.5, Claude Sonnet 5.5, DeepSeek V4 Pro, Gemini 2.5 Flash, Gemini 3.8 Flash, Gemini 4 Argon, Kimi K3, GPT-6.1 Sol, Qwen3.8 Max, Grok 4.7 and GLM-5.3, among others.

Connecting a client takes two settings: the address, and a key you create under API keys in the dashboard. The details live in the documentation, where they belong, and the address itself is something you copy rather than something you read on a marketing page. Once those two settings are in place, Gemini CLI talks to BestRelay in the ordinary way, and the model you typed is the model that does the work.

A relay earns its keep on small, verifiable properties like this one. BestRelay does not need to be large to be useful; it needs to be legible, and a short list of named models with published rates is a good deal more legible than an undifferentiated pool of endpoints.

Gemini models on BestRelay speak their own native protocol

Protocol questions are where relays tend to blur, so BestRelay draws the line plainly. Gemini models use their native protocol, reached through Gemini CLI or through direct generateContent calls. They are not OpenAI-compatible, and BestRelay does not pretend otherwise by wrapping one shape around another. If you work in Gemini CLI, the native route is the one you already know; if you write code against the provider, direct generateContent calls are available without translation.

The Claude side of the line-up follows a different, equally explicit rule. 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. Those tools, like Codex, pair with GPT, DeepSeek, Kimi, GLM, Qwen and Grok models instead.

Stated this way, the compatibility map is short enough to memorise, which is a form of value in itself. You are not left guessing whether a given client will accept a given model, and you are not left reading a support page that says everything works with everything. BestRelay publishes the boundaries of each route and lets you choose a combination that is actually supported.

For Gemini CLI users, the practical result is a clean path: native protocol, named model, published rates. For teams running more than one tool, the same table tells you where to put each client — Gemini CLI for Gemini, Claude Code or Claude Desktop for Claude, and the GPT, DeepSeek, Kimi, GLM, Qwen or Grok models for the general-purpose clients.

BestRelay publishes Gemini rate cards per token, not per seat

Every model on the BestRelay line-up carries four prices per 1,000,000 tokens: input, output, cache write and cache read. For the Gemini family, the cards read as follows. Gemini 2.5 Flash is $0.2778 input, $2.08 output, $0.2778 cache write and $0.0264 cache read. Gemini 3.1 Pro is $1.74 input, $9.93 output, $1.74 cache write and $0.1389 cache read. Gemini 3.8 Flash is $1.25 input, $5.56 output, $1.25 cache write and $0.1236 cache read. Gemini 4 Argon is $1.67 input, $8.06 output, $1.67 cache write and $0.0847 cache read.

Because the rates are per token rather than per seat, the arithmetic of everyday use is something you can do before you commit a budget. Take GPT-6.1 Sol as an example: a day of 500,000 input tokens 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 are $44.00 and the cached day is $1.33 — a difference of $1.78 on the day and $39.26 over the 22 days.

That example is not the only line of value in the table, but it shows the shape of it: real token volumes, real output lengths, and a rate card that can be read at a glance before the first call is made. The new-account credit of $0.35 covers about 1.6 million GPT-6.1 Sol input tokens, which is enough to see the service behave rather than guess at it.

Choosing between Gemini 2.5 Flash and Gemini 3.1 Pro becomes a question about the work rather than about hidden routing. A cheap card for routine drafting, a pricier card for heavier reasoning, and a published number for each. That is a modest feature, and it is the sort of modesty that keeps a bill predictable.

BestRelay bills once, after the call, from a prepaid balance

BestRelay uses 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, and there is no second meter running quietly beside the first.

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 relay that bills on the provider's own token report and declines to charge for a call that did not happen is doing two useful things at once: it keeps the accounting honest and it removes the argument.

Top-ups go through Stripe checkout in the amounts $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89, so you can size the balance to the week rather than committing to a fixed plan. 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.

There is also a referral arrangement worth reading carefully. 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. As an example, on an invited member's top-up of $13.89, the two commissions are $2.50 and $0.5556. Commission can be moved into the balance, or withdrawn to Alipay from $20.83. It is a small, specific arrangement rather than a promise of passive income, and it fits comfortably inside a pay-as-you-go account you control.

BestRelay keeps Gemini CLI keys, usage records and privacy in plain view

An account can hold several keys, and each can be voided at any time, which makes it straightforward to separate a client, a project or a colleague without disturbing the rest of the account. Limits are stated in the same plain register: 120 requests a minute on one key, and 8 calls in flight on one account. No ambiguity about whether a burst will be absorbed or refused.

Every call appears in the usage records with its model, tokens and charge. For Gemini CLI work, that means you can confirm that the request you sent to Gemini 3.8 Flash was answered by Gemini 3.8 Flash, and you can see the tokens the provider reported and the amount deducted. The record and the bill are the same document, which removes a category of small worry that would otherwise accumulate over a busy month.

On privacy, the terms are specific. 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. The retention is bounded, its purpose is stated, and it applies to the part of the request you actually wrote.

Put together, the picture is a relay that prefers plain statements to grand ones: a curated line-up of 20 public models, native protocol for Gemini through Gemini CLI, rate cards in the open, one charge per completed call, and usage records you can read. If that arrangement suits the way you work, BestRelay is ready when you are, and the $0.35 opening credit is a modest way to begin.