BestRelay API relay: Claude Opus 5.5 pricing at $1.74 per 1M

BestRelay forwards every request to the model name you wrote

An example day at your own desk. You open the BestRelay dashboard, create a key, and begin with a model you have chosen yourself. BestRelay is an independent API relay, and it forwards each request to the model provider under the model name you wrote. The model named is the model that answers, and nothing is quietly swapped for a cheaper tier. It is worth stating that plainly, because a relay that conceals its routing makes it hard to know what you are actually paying for.

For example, suppose your day involves drafting in Claude Opus 5.5, a short lookup in GPT-6.1 Sol, and a passing question for Grok 4.7. You name each one as you write the request. BestRelay passes those names through to the provider, so the replies come from the models you asked for rather than from a house substitute wearing the same label.

The line-up is curated rather than sprawling: 20 public models, each listed under its own name. When a provider retires a model id, BestRelay delists it. A retired id is not quietly kept alive under an old heading, and it is not replaced by another model behind your back. What the list says is what you can call.

The rest of the example day follows the same pattern. You write, the request leaves under the name you supplied, and the usage records later show the model, the tokens and the charge. None of this asks for a monthly commitment or a contract. BestRelay is a relay service built around one routing rule, and the rule is simple enough to check at a glance: name the model, get that model.

BestRelay lists 20 curated models, each with four prices per 1,000,000 tokens

Every model in the BestRelay line-up carries its own rate card, with four prices where the provider charges four: input, output, cache write and cache read. Models such as DeepSeek V4 Pro and Kimi K3 list input, output and cache read. The figures are published per 1,000,000 tokens, so you can compare a day of work against a month of estimates without arithmetic gymnastics.

For example, Claude Sonnet 5.5 is listed at $1.39 input and $7.01 output, with cache write at $1.74 and cache read at $0.1222. GPT-6.1 Sol is listed at $0.2083 input and $1.11 output, with cache read at $0.0431. Gemini 3.1 Pro shows $1.74 input and $9.93 output. Qwen3.8 Max shows $0.8333 input and $2.57 output, and GLM-5.3 shows $0.625 input and $1.94 output.

For example, a day of 500,000 input and 100,000 output tokens on Claude Sonnet 5.5 costs $1.40 on BestRelay, and 22 such days cost $30.71; with 350,000 of the input read from cache, the day costs $0.95. At the official list price the same day is $3.00, the 22 days $66.00 and the cached day $2.06, a difference of $1.60 on the day and $35.29 over the 22 days.

For example, the same shape of day on GPT-6.1 Sol costs $0.22 on BestRelay, with 22 such days at $4.74 and a cached day at $0.16. At the official list price that day is $2.00, the 22 days $44.00 and the cached day $1.48, differences of $1.78 and $39.26. BestRelay publishes both sides of that comparison, which is how you decide which model fits the work.

BestRelay bills a prepaid pay-as-you-go balance after each call ends

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, using the tokens the provider reports and the rate card prices for the model you named. Nothing is held in advance, so an idle balance is not quietly bleeding.

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 did not run is not a bill.

Every new account opens with $0.35 of credit, which is enough to start. For example, that free credit alone covers about 1.6 million GPT-6.1 Sol input tokens. Top-ups go through Stripe checkout in these amounts: $1.39, $6.94, $13.89, $27.78, $69.44 and $138.89, so you can add a little or add a month at a time.

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 simple to state: 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. For example, on an invited member's top-up of $13.89, the two commissions are $2.50 and $0.5556. Everything about the money sits in one place, which is the sort of dull clarity a working desk can use.

BestRelay sets clear client rules for Claude models and for Gemini models

Not every model suits every tool, and BestRelay says so rather than leaving you to find out. 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. Those tools, like Codex, pair only with GPT, DeepSeek, Kimi, GLM, Qwen and Grok models.

Gemini models use only their native protocol, through Gemini CLI or direct generateContent calls, and are not OpenAI-compatible. A Gemini 3.8 Flash or Gemini 4 Argon request therefore belongs in Gemini CLI or a direct generateContent call, rather than in an OpenAI-shaped client.

For example, in the example day, the Claude Opus 5.5 drafting work happens in Claude Code, and the GPT-6.1 Sol lookup runs in Codex. Each model meets a tool that speaks its protocol, and the Gemini 3.1 Pro question is handled separately, in Gemini CLI, on its own native route.

Connecting a client takes two settings: the address and a key created under API keys in the dashboard. The documentation holds the details, and the address itself is not printed here. The setup is the same small procedure whichever model you choose, which keeps the example day from turning into an afternoon of configuration.

BestRelay keeps keys, usage records and limits under your control

An account can hold several keys, each of which can be voided at any time. That makes it easy to retire a key when a machine changes hands, and it keeps one stray script from holding a credential indefinitely.

Limits are stated in numbers you can plan around: 120 requests a minute on one key, and 8 calls in flight on one account. For example, a long-running queue on one key stays inside the per-minute allowance, and a burst of parallel work on one account stays inside the in-flight limit.

Every call appears in the usage records with its model, tokens and charge, which makes the day-end check a short read rather than an investigation. The $0.35 opening credit, the $1.39 top-up and the $13.89 top-up all sit in the same ledger, so the arithmetic on your own account is visible to you.

On privacy, the position is narrow on purpose: the user-written text of requests is kept for 30 days to review abuse and settle billing questions, while system prompts and model replies are not stored.

That closes the example day. You named the models, the models answered, the calls were billed once, and the records show it. If that arrangement suits the way you work, BestRelay is ready when you are.