Grok 4.7 on BestRelay: a sample day through an API relay

Model portraitsEN2026-11-02

BestRelay forwards Grok 4.7 to the provider under the name you write

Here is an example day, plainly labelled as an example: you are at your own desk, a client is open, and you ask for Grok 4.7. The reply that comes back comes from Grok 4.7. That is the whole promise BestRelay makes, and it is worth stating precisely rather than loosely. BestRelay is an independent API relay. It forwards each request 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 behind the curtain. If your request says Grok 4.7, Grok 4.7 answers.

This matters more than it sounds. A relay that quietly reroutes work can produce results that look fine and are quietly wrong, and you have no obvious way to tell from the reply alone. BestRelay's arrangement removes that uncertainty by construction: the name you write is the name used for the forward. The same rule holds across the line-up. Ask for Claude Opus 5.5 and Claude Opus 5.5 answers; ask for GPT-6.1 Sol, Gemini 3.1 Pro, Kimi K3, Qwen3.8 Max, GLM-5.3 or DeepSeek V4 Pro, and the provider receives that name.

On the example desk, that makes the routine unremarkable in a workmanlike way. You write the model name, you send the request, you read the answer. You do not negotiate with a router, choose a tier, or guess whether a label describes the model underneath. BestRelay is an independent relay that passes your request along faithfully, and its curated line-up is where you choose the name to begin with. Nothing about the arrangement asks you to take the mapping on trust; the usage records show the model for every call, so the desk ledger and the invoice agree with each other.

Grok 4.7 sits in a curated line-up of 20 public models, each with its own prices

The line-up is curated and finite: 20 public models, each listed under its own name with four prices — input, output, cache write and cache read — so you can compare like with like before committing a token budget. Grok 4.7 appears as Grok 4.7, with input $0.625, output $1.94 and cache read $0.1389 per 1,000,000 tokens. Grok 4.6 sits beside it at input $0.625 and output $1.81, and GLM-5.3 carries the same $0.625 and $1.94 figures, so the choice between them turns on the work itself rather than on how the listing happens to be worded.

Other corners of the line-up read differently. Gemini 2.5 Flash lists input $0.2778 and output $2.08; Gemini 3.8 Flash lists input $1.25 and output $5.56; Gemini 4 Argon lists input $1.67 and output $8.06; Gemini 3.1 Pro lists input $1.74 and output $9.93. Claude Fable 5 lists input $4.17 and output $20.97, while Claude Haiku 4.5 lists input $0.4167 and output $2.01. GPT-6.1 Sol lists input $0.2083 and output $1.11, and GPT-5.6 Sol lists input $0.4167 and output $2.29. Every one of these is a separate entry with its own rate card, not a tier concealed behind a friendly label.

The catalogue is also kept current over time. When a provider retires a model id, BestRelay delists it, so a name that has gone away does not linger in the list as a way to spend money on nothing. That discipline is what makes a curated line-up worth reading in the first place: twenty entries, each with prices you can read before you type, each pointing at a model that actually answers. Reading the list is also the fastest way to see what the desk day might cost, since the four prices for each name are all printed in the same form.

An example desk day on Grok 4.7 costs $0.51 on BestRelay

Here is the example desk again, with numbers attached. As an example, a day of 500,000 input tokens and 100,000 output tokens on Grok 4.7 costs $0.51 on BestRelay, and 22 such days come to $11.15. If 350,000 of that day's input is read from cache instead, the same day costs $0.34. These are examples of arithmetic on the published rate card, not a quotation or a forecast of your own use.

The shape of that figure is the point. Caching is priced openly — for Grok 4.7, cache read is $0.1389 per 1,000,000 tokens — so a workload that repeats itself does not pay the full input price for text the provider has already seen. If your desk day involves a long standing brief you send again each morning, the cached figure is the more relevant number to keep in mind. The rate card also lists cache write as a separate line, so a workflow that stores context can be accounted for rather than guessed at.

For comparison, the same style of example on other entries: Claude Sonnet 5.5 costs $1.40 for a day of 500,000 input and 100,000 output tokens, $30.71 over 22 such days, and $0.94 when 350,000 of the input is read from cache. GPT-6.1 Sol comes to $0.22 for the day, $4.74 over 22 days and $0.16 with the cached input. The point is not that one model suits you, but that BestRelay lets you price the day before you start it, and lets you switch names within the same line-up without changing how the bill is calculated.

BestRelay bills prepaid pay-as-you-go, and failed calls are recorded at zero

Billing on BestRelay is prepaid pay-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 for the model you named. Nothing is held in advance beyond the balance you have already chosen to put there, and each new account opens with $0.35 of credit.

Failure is handled without drama. 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 does not quietly become an invoice line. The $0.35 opening credit alone covers about 1.6 million GPT-6.1 Sol input tokens, which is enough to sit with the service for a while before topping up.

Top-ups go through Stripe checkout, in these amounts:

$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.

Referrals pay 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.

Connecting a client to BestRelay takes an address and a key from the dashboard

A client needs two settings: the address, and a key created under API keys in the dashboard. The details are in the docs, and the address itself is not something to print in a blog post, a screenshot or a chat message. Create the key, enter the two settings in the client, and the model names in the line-up become available to you.

Grok 4.7 is a natural companion for tools such as Codex, Cursor, Cherry Studio, Cline, OpenCode, any SDK or cURL. Claude models are served only to the official Claude Code and Claude Desktop clients, version 2.1.285 or newer, and they do not work in Cursor, Cherry Studio, Cline, OpenCode, any SDK or cURL. Gemini models use only their native protocol, through Gemini CLI or direct generateContent calls, and are not OpenAI-compatible.

Pace and bookkeeping are equally plain. One key allows 120 requests a minute, and one account allows 8 calls in flight. An account can hold several keys, each can be voided at any time, and every call appears in the usage records with its model, tokens and charge. Your request text is kept for 30 days to review abuse and settle billing questions, while system prompts and model replies are not stored.

If the example desk day sounds like your own, BestRelay is worth a try with your own key and your own model names.