BestRelay pay-as-you-go API relay for Sonnet 5.5 users

BestRelay answers with the model you named

Here is an example of a working day, drawn as a sample rather than a claim about you. You sit down with a coffee and a small backlog: a support macro, a pricing page, a half-written migration note. You open your preferred client, type the model name you want, and BestRelay forwards that request to the provider under the name you wrote. The model that answers is the model you named. Nothing is quietly swapped for a cheaper tier on the way through, which is the whole point of an independent relay and the reason BestRelay keeps a curated line-up rather than an open catalogue. Later, when a provider retires a model id, BestRelay delists it, so the name on your rate card is a name you can actually call. Across the day, you use a Claude model for the long-form note, a Gemini model for a quick classification, and a GPT model for a code review. The arithmetic is not abstract: on the BestRelay rate card, Gemini 2.5 Flash is input $0.2778 and output $2.08 per 1,000,000 tokens, while Claude Opus 5.5 is input $1.74 and output $8.13. BestRelay does not care which family you prefer; it cares that the request reaches the provider you selected, with the pricing shown for that model. The rest of this example follows what that day looks like in money, in records, and in the small administrative choices that keep a young startup's model spend legible.

BestRelay lists 20 public models with four prices each

Every model in the BestRelay line-up is listed under its own name with four prices: input, output, cache write and cache read. There are 20 public models, and the rate card shows the numbers per 1,000,000 tokens before you commit a cent. Claude Fable 5 is listed at input $4.17, output $20.97, cache write $5.14 and cache read $0.3472, against an official list price of input $10, output $50 and cache read $1. Claude Opus 5.5 sits at input $1.74, output $8.13, cache write $2.01 and cache read $0.0778, against official list prices of input $4, output $20 and cache read $0.2. The point of listing all four prices is that cache behaviour changes the arithmetic of a day, so hiding it would be unhelpful. Take Claude Sonnet 5.5, for example: a day of 500,000 input and 100,000 output tokens 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, so the difference on the day is $0.60 and on the 22 days $13.29. GPT-6.1 Sol tells the same story more sharply: as an example, a day of 500,000 input and 100,000 output tokens costs $0.22 on BestRelay, 22 such days cost $4.74, and with 350,000 of the input read from cache the day costs $0.16, against $2.00, $44.00 and $1.33 at official list price, a difference of $1.78 on the day and $39.26 over the 22 days. The $0.35 of free credit each new account opens with covers about 1.6 million GPT-6.1 Sol input tokens.

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

BestRelay bills on a prepaid pay-as-you-go balance: no monthly fee, no subscription and no minimum spend. You top up through Stripe checkout in fixed amounts — $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89 — and the money is then drawn down as you use it. Each call is charged once, after it ends, from the tokens the provider reports, at the rate card prices for the model that ran. Nothing is held in advance, which means the balance you see is the balance you have. Failure has a plain rule attached to it: requests the provider rejects are recorded at zero, and when the provider cannot be reached nothing is charged. A rejected request is not a small loss to absorb quietly; it is a line in the usage records at zero. Refunds are handled in plain terms too. Unused credit can be refunded on request within 13 days of the top-up, less what was consumed and any payment-processor fees, while gift credit and commission cannot be refunded. The referral scheme is worth noting for a young team introducing the service to friends: 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. To take 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.

BestRelay pairs each model with the clients that can reach it

The model you name is only half of a working setup; the client is the other half, and BestRelay is careful about the pairing. 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, and the same goes for Codex: those tools 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 they are not OpenAI-compatible. That last point saves a common afternoon of confusion, when a request fails because the shape of the call is wrong rather than the model. Connecting a client takes two settings, the address and a key created under API keys in the dashboard; the details are in the BestRelay docs, and they are deliberately not reprinted here, because a printed address is a printed address. The rule of thumb for your sample day is simple. If you are drafting a migration note in Claude Code, the model you type is the model that answers. If you are classifying support tickets through Gemini CLI, the same holds. If you are reviewing code in Codex, you are working in the GPT, DeepSeek, Kimi, GLM, Qwen and Grok part of the line-up, and BestRelay lists each of those models under its own name with its own four prices.

BestRelay keeps keys, records and limits in plain sight

Control at the end of the day comes down to records, keys and a few hard limits. A BestRelay account can hold several keys, each of which can be voided at any time, and every call appears in the usage records with its model, tokens and charge. That is the sentence a founder actually wants at 6 p.m.: which model ran, how many tokens it used, and what it cost. The limits are stated plainly too. One key is good for 120 requests a minute, and an account can have 8 calls in flight. Neither figure is a suggestion; they are the shape of the service, and knowing them in advance is kinder than discovering them mid-demo. On privacy, BestRelay keeps the user-written text of requests for 30 days so that abuse can be reviewed and billing questions settled; system prompts and model replies are not stored. So the copy you send is the copy that is retained, briefly and for stated reasons, and the model's answers are not kept. Put the day together, as an example: you start with $0.35 of credit, top up through Stripe checkout, work through Claude and Gemini and GPT models in the clients that support them, watch each call land in the usage records with its model, tokens and charge, and end the day with a balance that reflects tokens actually reported by the provider. Nothing was held in advance, nothing was swapped on the way through, and nothing about the spend was hidden from you. If that arrangement suits how you like to work, BestRelay is ready to sit on your desk the same way.