BestRelay pricing for Claude Opus 5.5, read line by line

BestRelay forwards each request to the model the caller names

BestRelay is an independent API relay, and the promise it makes is narrow enough to check on every call: the model name you write is the model name that answers. A request that says Claude Opus 5.5 is forwarded to the provider under Claude Opus 5.5. Nothing is quietly swapped for a cheaper tier behind the scenes. BestRelay sits between your client and the provider, and it leaves the choice of model in your hands rather than in a hidden routing table.

The line-up is curated rather than exhaustive. Twenty public models are listed, each under its own name, and each carries four prices: input, output, cache write and cache read. When a provider retires a model id, BestRelay delists it rather than leaving a name on the card that no longer resolves. That means the card you read is the card you can use, and a name you choose is a name you can point at later in the usage records.

For teams, this matters more than it might appear. A shared project often runs several models at once, a lighter one for routine work and a heavier one for the difficult parts, and the temptation with an unnamed relay is that the label on the invoice stops describing the work being done. BestRelay closes that gap between label and response. You name the model, you get the model, and the charge follows the same name.

That arrangement also makes a budget conversation easier. Instead of arguing about which tier your traffic quietly fell into, you can read the model field in the records and know exactly which line of the rate card produced the figure. The rest of this article takes that card apart line by line, then looks at how the balance, the clients and the records fit around it.

Twenty public models, each with four prices per million tokens

Reading the BestRelay rate card is not difficult, because the format never varies. Every entry is a model name followed by four figures per 1,000,000 tokens: input, output, cache write and cache read. Some models publish a cache write price and some do not, and where a figure is absent the card simply does not carry it.

Take the Claude entries. Claude Fable 5 is priced at $4.17 input, $20.97 output, $5.14 cache write and $0.3472 cache read, against an official list price of $10 input, $50 output and $1 cache read. Claude Haiku 4.5 sits lower at $0.4167 input, $2.01 output, $0.4861 cache write and $0.0333 cache read. Claude Opus 5.5 lists $1.74 input, $8.13 output, $2.01 cache write and $0.0778 cache read. Claude Sonnet 5.5 lists $1.39 input, $7.01 output, $1.74 cache write and $0.0944 cache read.

Elsewhere the card moves between families. GPT-6.1 Sol is $0.2083 input, $1.11 output and $0.0431 cache read. GPT-5.6 Sol is $0.4167 input, $2.29 output and $0.0333 cache read. Gemini 3.1 Pro is $1.74 input, $9.93 output, $1.74 cache write and $0.1389 cache read. Gemini 2.5 Flash is $0.2778 input, $2.08 output, $0.2778 cache write and $0.0264 cache read. DeepSeek V4 Pro is $1.53 input, $4.58 output and $0.0514 cache read. Kimi K3 is $1.39 input, $7.01 output and $0.1389 cache read. Qwen3.8 Max is $0.8333 input, $2.57 output and $0.1014 cache read. GLM-5.3 and Grok 4.7 are both $0.625 input, $1.94 output and $0.1389 cache read, and Grok 4.6 is $0.625 input, $1.81 output and $0.1389 cache read. Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5 and GPT-6 Astra complete the card.

Two details are worth noticing as you read. First, the four columns are not decorative: input and output are priced separately because the work differs, and cache write and cache read are priced separately because one stores and the other retrieves. Second, the card stays short on purpose. A line-up of twenty public models is a line-up someone has chosen, and a name that has been retired is removed rather than kept as a ghost entry.

BestRelay charges once per finished call, from a prepaid balance

BestRelay bills the way a careful accountant would prefer to read: one charge per call, taken after the call ends, from a balance you have already paid for. There is no monthly fee, no subscription and no minimum spend. The amount comes from the tokens the provider reports, priced at the rate card figures above. Nothing is held in advance, so a request that never finishes bills nothing.

The prepaid balance is topped up through Stripe checkout in fixed amounts: $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89. A small top-up is enough to try a model; a larger one suits a team running the same workload every day. Each new account opens with $0.35 of credit, which is enough to make a first call and watch the records populate.

Failures are treated plainly. Requests the provider rejects are recorded at zero, and when the provider cannot be reached at all, nothing is charged. The usage records therefore show the work and the absence of work, rather than a single blended number that leaves you guessing where the money went.

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. The rule is short enough to hold in your head and useful enough to keep in writing: pay for what you use, and ask for the remainder back if you overfilled the balance early on. For a company that wants a readable bill, this combination of prepaid funds, one charge per call and a stated refund window is about as plain as a billing arrangement gets.

Claude Code and Gemini CLI need different routes to the line-up

Setting up BestRelay takes two settings on the client side: the address and a key created under API keys in the dashboard. The details live in the documentation. That is all the configuration a client needs, and it is worth reading carefully, because the models on this card do not all speak the same protocol.

Claude models are served to the official Claude Code and Claude Desktop clients, version 2.1.285 or newer. Other clients, including Cursor, Cherry Studio, Cline and OpenCode, pair with GPT, DeepSeek, Kimi, GLM, Qwen and Grok models, and Codex works with that same family of names. Gemini models use their native protocol, reached through Gemini CLI or direct generateContent calls. Choosing the right client for the model you named is the whole of the setup; the rest is the two settings.

The limits are stated in plain numbers. One key handles 120 requests a minute, and one account runs 8 calls in flight. For a team, that shape suits interactive work and scripts that fire requests in sequence, and it is clear enough to plan around without guesswork.

An account can hold several keys, so a build pipeline, a colleague and a local script do not have to share one credential. Any key can be voided at any time. Each call then appears in the usage records with its model, tokens and charge, which is the record you would want when a budget question arrives and someone asks what a particular afternoon cost. 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.

Usage records, referrals and a bill that can be read line by line

Two examples show how the card behaves in ordinary daily use. Take Claude Sonnet 5.5: as an 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; the difference on the day is $0.60 and on the 22 days $13.29.

Take GPT-6.1 Sol as an example: the same 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. 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.

Referrals are paid at 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.

That is the shape of BestRelay: named models, a rate card you can read line by line, and a bill that says what it should. If you would like to see how that behaves on your own workload, start with the free credit.