BestRelay API relay setup with GPT-6.1 Sol and Claude Code

BestRelay bills each finished call once, from a prepaid balance

BestRelay is an independent API relay, and its billing is easy to read: there is no monthly fee, no subscription and no minimum spend. You open an account and it starts with $0.35 of credit. You then top up through Stripe checkout in fixed amounts — $1.39, $6.94, $13.89, $27.78, $69.44 or $138.89 — and that money waits on the balance until calls draw it down. Nothing is held in advance beyond what you have chosen to fund.

Each call is charged once, after it ends, using the token counts the provider reports and the rates printed on the card. The bill therefore follows the work you asked for, rather than a flat monthly figure that arrives whether you use the service or not. A quiet month costs you nothing to keep the account running; a busy one costs what it used. Requests the provider rejects are recorded at zero, and when the provider cannot be reached at all, nothing is charged.

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, which is worth knowing before treating a referral balance as though it were a top-up. For anyone who reads a rate card before committing, the arrangement is plain: put in a small amount, watch what it buys, and decide later. BestRelay asks for no standing commitment, and untouched credit can be asked back for inside that window.

BestRelay's rate card is a list of model names and published rates

The line-up is curated and holds 20 public models, each listed under its own name with rates per 1,000,000 tokens for input, output and cached work. When a provider retires a model id, BestRelay delists it, so the card does not accumulate names that no longer answer. Here are some of the lines as published, read in order: input, output, cache write, cache read.

• Claude Fable 5 — in $4.17, out $20.97, write $5.14, read $0.3472
• Claude Fable 5.1 — in $4.10, out $20.42, write $5.00, read $0.2083
• Claude Haiku 4.5 — in $0.4167, out $2.01, write $0.4861, read $0.0333
• Claude Opus 5 — in $2.29, out $11.46, write $2.85, read $0.1389
• Claude Opus 5.5 — in $1.74, out $8.13, write $2.01, read $0.0778
• Claude Sonnet 5 — in $1.32, out $6.67, write $1.67, read $0.1167
• Claude Sonnet 5.5 — in $1.39, out $7.01, write $1.74, read $0.0944
• Gemini 2.5 Flash — in $0.2778, out $2.08, write $0.2778, read $0.0264
• Gemini 3.1 Pro — in $1.74, out $9.93, write $1.74, read $0.1389
• Gemini 3.8 Flash — in $1.25, out $5.56, write $1.25, read $0.1236
• Gemini 4 Argon — in $1.67, out $8.06, write $1.67, read $0.0847
• GPT-6.1 Sol — in $0.2083, out $1.11, read $0.0431

The remaining lines follow the same format. DeepSeek V4 Pro is $1.53 in, $4.58 out, $0.0514 cache read. Kimi K3 is $1.39 in, $7.01 out, $0.1389 cache read. Qwen3.8 Max is $0.8333 in, $2.57 out, $0.1014 cache read. Grok 4.6 is $0.625 in, $1.81 out, $0.1389 cache read; Grok 4.7 is $0.625 in, $1.94 out, $0.1389 cache read; GLM-5.3 is $0.625 in, $1.94 out, $0.1389 cache read. GPT-5.6 Sol is $0.4167 in, $2.29 out, $0.0333 cache read, and GPT-6 Astra is $0.7639 in, $3.96 out, $0.0792 cache read. Where the sheet records an official list price, it sits beside the BestRelay rate: Claude Fable 5 lists $10 input, $50 output and $1 cache read; GPT-6.1 Sol lists $2 input, $10 output and $0.1 cache read; Claude Opus 5.5 lists $4 input, $20 output and $0.2 cache read.

The model you name on BestRelay is the model that answers

BestRelay forwards each request to the model provider under the model name the caller wrote. The model named is the model that answers, and nothing is quietly swapped for a cheaper tier behind the scenes. That is the whole point of the curated line-up: you read a name on the card, you write that name in your request, and the provider receives it under those words. The rate you pay is the rate printed beside that name.

Client rules are stated just as plainly. Claude models are served to the official Claude Code and Claude Desktop clients, version 2.1.285 or newer, and to no other client. Tools such as Codex pair with GPT, DeepSeek, Kimi, GLM, Qwen and Grok models. Gemini models use their own native protocol, through Gemini CLI or direct generateContent calls, and they keep that protocol rather than being reshaped into something else. Knowing these boundaries before you build saves an afternoon of mismatched configuration.

Connecting a client takes two settings: the address and a key created under API keys in the dashboard. The details are written up in the docs. Beyond that, BestRelay keeps the record in the open: an account can hold several keys, any key can be voided at any time, and every call appears in the usage records with its model, tokens and charge. One key is limited to 120 requests a minute, and one account to 8 calls in flight. Request text you write is kept for 30 days to review abuse and settle billing questions; system prompts and model replies are not stored. BestRelay is careful about what it holds and about what it hands back.

Worked examples show what a day of traffic costs on BestRelay

Take Claude Sonnet 5.5, as an example. A day of 500,000 input and 100,000 output tokens costs $1.40 on BestRelay, and 22 such days cost $30.71. With 350,000 of the input read from cache, the day costs $0.94. 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. Nothing here is a promise about your own traffic. It is the card doing arithmetic, so that the arithmetic is not a surprise later.

Now take GPT-6.1 Sol, as an example. A day of 500,000 input and 100,000 output tokens costs $0.22 on BestRelay, and 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. The free credit alone covers about 1.6 million GPT-6.1 Sol input tokens, which is a reasonable way to test the relay before topping up.

These are examples, not averages. Your own mix of input, output and cached tokens decides the total, and BestRelay shows the tokens and the charge for each call so you can read the pattern in your usage records rather than infer it. A prepaid balance means the meter is the meter: whatever the card says, that is what leaves the balance when the call ends.

Setting up BestRelay takes an address, a key and a decision you can revisit

Two settings connect a client to BestRelay: the address and a key from API keys in the dashboard, with the details in the docs. Create the key, set the limits you want, and your calls are recorded as they run. If a key should stop working, void it; if you want to see what you spent, the usage records list every call with its model, tokens and charge. There is no monthly fee to cancel and no minimum spend to justify, so the decision to keep using BestRelay can be made again each month against real numbers.

BestRelay also pays for the word of mouth. 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. For 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. Gift credit and commission cannot be refunded, so treat that stream as separate from your top-up balance and keep the 13-day window in mind for the credit you funded yourself.

Read the card, pick a model, fund a small balance, and watch the first call appear in the usage records with its model, tokens and charge. BestRelay will be here when you come back with more.