BestRelay API relay for GLM-5.3 at $0.625 input

BestRelay sends your request under the exact model name you wrote

You sit down to translate a chapter or draft an essay, and the first thing you want from your tools is simple: the model you name is the model that answers. BestRelay is built around that sentence. BestRelay is an independent API relay, and it forwards each request to the model provider under the model name the caller wrote, so nothing is quietly swapped for a cheaper tier behind your back. That single rule is the reason the service feels predictable at a working desk, where a half-finished paragraph is often more expensive to redo than to pay for properly.

The line-up on BestRelay is curated rather than scraped together. It holds 20 public models, and each one is listed under its own name with four prices: input, output, cache write and cache read. When a provider retires a model id, BestRelay delists it, so the catalogue you read is the catalogue that exists on the day you read it. Nothing lingers as a name you can call and hope for.

For a translator or writer working with GLM-5.3 and Kimi K3, that arrangement matters in a quiet, practical way. Both models appear on BestRelay under their own names, each with its own rate card, and a request that says Kimi K3 is forwarded to Kimi K3. The same goes for GLM-5.3. You are not buying an allowance of some unnamed engine and hoping the tone comes out right; you are choosing a named model and seeing that name in the usage records afterwards, beside its tokens and its charge.

A sample day at your desk with GLM-5.3 and Kimi K3, as an example

Here is a sample day at your own desk, plainly labelled as an example. In the morning you run GLM-5.3 over long source passages; in the afternoon you ask Kimi K3 to polish the English. On BestRelay, GLM-5.3 is priced per 1,000,000 tokens at $0.625 input, $1.94 output and $0.1389 cache read, and Kimi K3 at $1.39 input, $7.01 output and $0.1389 cache read. Those are the figures you work from, and they are the figures the service bills against.

As an example, a day of 500,000 input and 100,000 output tokens on GLM-5.3 costs $0.51 on BestRelay, and 22 such days cost $11.15; with 350,000 of the input read from cache the day costs $0.34. As an example, the same day on Kimi K3 costs $1.40 on BestRelay, 22 such days cost $30.71, and with 350,000 of the input read from cache the day costs $0.96. The cached figures are the ones a translator working through the same manuscript twice a week will notice, because the second pass reads a good share of its context back from cache.

For comparison, as an example, a day of 500,000 input and 100,000 output tokens on Claude Sonnet 5.5 costs $1.40 on BestRelay, 22 such days cost $30.71, and 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. Claude Sonnet 5.5 is priced on BestRelay at $1.39 input, $7.01 output, $1.74 cache write and $0.0944 cache read, which is where those totals come from.

Put the two days side by side and the choice becomes a working question rather than a slogan. GLM-5.3 at $0.625 input suits the heavy reading side of a translation desk; Kimi K3 at $1.39 input suits the rewriting side where the output tokens matter. Either way, the model named in the client is the model that answers, and the usage records show you which one produced which passage.

BestRelay bills pay-as-you-go from a prepaid balance with no monthly fee

Billing on BestRelay is a prepaid pay-as-you-go balance: no monthly fee, no subscription, no minimum spend. Each call is charged once, after it ends, from the tokens the provider reports, at the rate card prices, and nothing is held in advance. For someone who writes in bursts, that is a comfortable arrangement, because a quiet fortnight costs a quiet balance rather than an unused standing order.

Failed calls are treated plainly. Requests the provider rejects are recorded at zero, and when the provider cannot be reached nothing is charged. A timeout at eleven at night therefore does not quietly become an invoice line, which matters when you are the one who chose the model and set the budget.

Each new account on BestRelay opens with $0.35 of credit; as an example, the free credit alone covers about 1.6 million GPT-6.1 Sol input tokens, which is enough to try the service on a real paragraph 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 are handled in the open 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. 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, so a small share of someone else's top-up either becomes working credit or becomes money you can take out.

GLM-5.3 and Kimi K3 pair with the tools you already use on BestRelay

On BestRelay, the client rules are plain and worth reading before you configure anything. 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. Those tools, like Codex, pair only with GPT, DeepSeek, Kimi, GLM, Qwen and Grok. That is where GLM-5.3 and Kimi K3 belong, so both fit the editors and coding clients a writer already keeps open.

Gemini models use only their native protocol, through Gemini CLI or direct generateContent calls, and are not OpenAI-compatible. The Gemini entries on BestRelay are priced per 1,000,000 tokens as well: Gemini 2.5 Flash at $0.2778 input, $2.08 output, $0.2778 cache write and $0.0264 cache read; Gemini 3.1 Pro at $1.74 input, $9.93 output, $1.74 cache write and $0.1389 cache read; Gemini 3.8 Flash at $1.25 input, $5.56 output, $1.25 cache write and $0.1236 cache read; and Gemini 4 Argon at $1.67 input, $8.06 output, $1.67 cache write and $0.0847 cache read.

To connect a client to BestRelay, you need two settings: the address and a key created under API keys in the dashboard. The details are in the docs, and we do not print the address or any setting here. The key is yours to name, and several keys can sit under one account, which is useful when you want a separate key for drafting and a separate key for the finished translation.

Limits on BestRelay are set per key and per account rather than hidden behind a queue. One key allows 120 requests a minute, and one account allows 8 calls in flight at once. For a desk that runs one model over a manuscript while another polishes a section, that shape is workable, and you can see each call afterwards in the usage records with its model, tokens and charge.

Keys, usage records and privacy on BestRelay

Keeping track of a desk is easier when the paperwork is automatic. An account on BestRelay 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. If a key is ever mislaid or used where it should not have been, voiding it takes effect at once, and the remaining calls still report their own model and cost, so the arithmetic on a month of work stays legible.

The rest of the line-up is priced in the same way, which makes comparison less effort. DeepSeek V4 Pro is listed at $1.53 input, $4.58 output and $0.0514 cache read; Qwen3.8 Max at $0.8333 input, $2.57 output and $0.1014 cache read; Grok 4.6 at $0.625 input, $1.81 output and $0.1389 cache read; Grok 4.7 at $0.625 input, $1.94 output and $0.1389 cache read; and GPT-6.1 Sol at $0.2083 input, $1.11 output and $0.0431 cache read. GPT-5.6 Sol sits at $0.4167 input, $2.29 output and $0.0333 cache read, and GPT-6 Astra at $0.7639 input, $3.96 output and $0.0792 cache read.

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. The distinction is worth noting for a translator handling unpublished manuscripts: what you type is retained for a defined period and a stated reason, while the system prompt and the model's answer are not kept.

Put together, the relay behaviour, the pay-as-you-go balance, the client line-up and the records give a translator or writer a desk that behaves the same way every day, and the rate card tells you in advance what a day of work will cost. We invite you to try BestRelay.