BestRelay pay-as-you-go setup with GPT-6.1 Sol for small top-ups

BestRelay charges the model you name, at the rate card prices

A rate card is useful only if you can read it without guessing. BestRelay is an independent API relay, and its line-up is curated: 20 public models, each listed under its own name, each with four prices per 1,000,000 tokens — input, output, cache write and cache read. A request is forwarded 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 different tier.

Reading the Claude Fable 5 row: input $4.17, output $20.97, cache write $5.14, cache read $0.3472. Claude Haiku 4.5 reads input $0.4167, output $2.01, cache write $0.4861, cache read $0.0333. GPT-6.1 Sol reads input $0.2083, output $1.11, cache read $0.0431. Gemini 2.5 Flash reads input $0.2778, output $2.08, cache write $0.2778, cache read $0.0264. When a provider retires a model id, BestRelay delists it, so the card stays a record of what can be called now.

Billing follows the same plainness. BestRelay runs on 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. Nothing is held in advance. Requests the provider rejects are recorded at zero, and when the provider cannot be reached, nothing is charged.

That is the whole reading of the card: four numbers per model, one balance, one charge per completed call.

Small BestRelay top-ups start at $1.39 and suit a side project's uneven month

BestRelay suits the case where a side project uses a model now and then. Top-ups go through Stripe checkout in these amounts: $1.39, $6.94, $13.89, $27.78, $69.44, $138.89. The $1.39 step is enough to begin, and larger steps exist for a project whose traffic grows. There is no monthly fee, no subscription and no minimum spend, so a quiet month costs nothing beyond what the calls consume.

Each new account opens with $0.35 of credit. For example, the free credit alone covers about 1.6 million GPT-6.1 Sol input tokens, which makes it a reasonable way to see how a prompt shape behaves before adding money.

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.

BestRelay also credits the people who bring others in: 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.

For a project that runs on a trickle of tokens, that arithmetic is the point: the balance is topped up in small steps, and every call is paid for by the tokens it actually used.

The Claude rows on BestRelay read below the official list prices

Start with the comparison the sheet allows. Claude Fable 5 is $4.17 input, $20.97 output, $5.14 cache write and $0.3472 cache read, against an official list price of input $10, output $50, cache read $1. Claude Fable 5.1 is $4.10, $20.42, $5.00 and $0.2083, against input $10, output $50, cache read $0.25. Claude Haiku 4.5 is $0.4167, $2.01, $0.4861 and $0.0333, against input $1, output $5, cache read $0.1.

The mid rows follow the same shape. Claude Opus 5 is $2.29 input, $11.46 output, $2.85 cache write and $0.1389 cache read, against input $5, output $25, cache read $0.5. Claude Opus 5.5 is $1.74, $8.13, $2.01 and $0.0778, against input $4, output $20, cache read $0.2. Claude Sonnet 5 is $1.32, $6.67, $1.67 and $0.1167, against input $2, output $10, cache read $0.2. Claude Sonnet 5.5 is $1.39, $7.01, $1.74 and $0.0944, against input $2, output $10, cache read $0.1.

Worked example, Claude Sonnet 5.5: a day of 500,000 input and 100,000 output tokens costs $1.40 on BestRelay, 22 such days cost $30.71; 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.

Cache write and cache read are separate columns because they are separate costs, and BestRelay lists both rather than folding them into a single figure.

BestRelay lists Gemini, GPT, DeepSeek, Kimi, Qwen, Grok and GLM on their own rows

The rest of the card is equally legible. DeepSeek V4 Pro: input $1.53, output $4.58, cache read $0.0514. Gemini 3.1 Pro: $1.74, $9.93, $1.74, $0.1389. Gemini 3.8 Flash: $1.25, $5.56, $1.25, $0.1236. Gemini 4 Argon: $1.67, $8.06, $1.67, $0.0847. Kimi K3: input $1.39, output $7.01, cache read $0.1389. Qwen3.8 Max: input $0.8333, output $2.57, cache read $0.1014. Grok 4.6: input $0.625, output $1.81, cache read $0.1389. Grok 4.7: $0.625, $1.94, $0.1389. GLM-5.3: $0.625, $1.94, $0.1389.

The GPT rows read like this: GPT-5.6 Sol input $0.4167, output $2.29, cache read $0.0333, against an official list price of input $4, output $20, cache read $0.4. GPT-6 Astra: $0.7639, $3.96, $0.0792, against input $10, output $50, cache read $1. GPT-6.1 Sol: $0.2083, $1.11, $0.0431, against input $2, output $10, cache read $0.1.

Worked example, GPT-6.1 Sol: a day of 500,000 input and 100,000 output tokens costs $0.22 on BestRelay, 22 such days cost $4.74; 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.

A side project that runs a few hundred thousand tokens a day can therefore be budgeted from the card itself, line by line, without a subscription.

BestRelay clients, keys and records decide which model a request can reach

The rate card is half the service; the other half is how a request arrives. 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. Gemini models use only their native protocol, through Gemini CLI or direct generateContent calls, and are not OpenAI-compatible.

Connecting takes two settings: the address and a key created under API keys in the dashboard. The details are in the docs. BestRelay does not print the address or any setting in an article like this one, and neither should you paste them into a public place.

The limits are stated plainly too: 120 requests a minute on one key, and 8 calls in flight on one account. 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.

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. A side project can be small, but its records should not be vague, and the usage records show what each call was and what it cost.

BestRelay is an independent relay, and the model you name is the model that answers. If your side project has been waiting for a clean way to try one of these models without a monthly commitment, the rate card, the small top-ups and the $0.35 opening credit are a sensible place to start with BestRelay.