Claude Code pay-as-you-go: Relay365 duty desk pricing

What a quiet week actually costs on Relay365

A careful buyer rarely opens a pricing page by asking which model is the headline. The first question is more practical: what happens during the weeks when the side project sleeps? On Relay365, the answer is plain. A quiet week costs nothing beyond what you actually send. The service runs like a round-the-clock duty desk, and every call is settled on the token counts the upstream reports.

Relay365 is an independent third-party relay with a prepaid wallet and no monthly fee. There is no seat to keep alive and no subscription quietly billing an idle repository. If your weekend experiment sends nothing, nothing is charged for it. That is the whole point of pay-as-you-go for bursty work: the cost follows the traffic rather than the calendar.

Take Claude Fable 5, wire name claude-fable-5. Input is $3.79 per 1 million tokens and output is $18.94 per 1 million tokens. For example, if you send 20,000 input tokens and 5,000 output tokens a day, that is about $0.0758 of input plus about $0.0947 of output, roughly $0.1705 a day, and about $5.12 across 30 days. A project that runs for a week instead of a month simply runs for a week, and Relay365 bills only the tokens that actually moved.

Now take GPT-5.6 Sol, wire name gpt-5.6-sol, with input at $0.3736 and output at $2.09 per 1 million tokens. For example, if you send 50,000 input tokens and 20,000 output tokens a day, the day comes to about $0.0187 of input plus about $0.0418 of output, roughly $0.0605, or about $1.81 over 30 days. The official list price for GPT-5.6 Sol output is $20 per 1 million tokens, and the Relay365 output rate is about 10% of official.

Which token type is driving the bill

The next careful question is: what exactly am I paying for? On Relay365, every lane sits on one table and the four token types are priced separately. Input, output, cache write and cache read each carry their own rate, and every request can be traced in the usage log, so you can see which line item is actually moving.

For Claude Fable 5, cache write is $4.66 per 1 million tokens and cache read is $0.2917. The official list price for cache read is $1 per 1 million tokens, while the official list price for input is $10 and for output is $50, so the Relay365 output rate is about 38% of official. A buyer who caches aggressively should read those two cache lines separately from the input and output lines, because they are billed apart rather than folded into one blended figure.

Gemini 2.5 Flash shows the same structure from the other end of the table: input $0.2389, output $1.91, cache write $0.2389 and cache read $0.0236 per 1 million tokens. For example, if you send 20,000 input tokens and 5,000 output tokens a day, that is about $0.0048 of input plus about $0.0096 of output, roughly $0.0143 a day and about $0.43 across 30 days. No official price is quoted for this family, so the honest comparison here is simply between the lanes on the same Relay365 table.

MiniMax M3, wire name MiniMax-M3, is priced at input $0.1528, output $0.6111 and cache read $0.0306 per 1 million tokens. For example, if you send 30,000 input tokens and 10,000 output tokens a day, that is about $0.0046 of input plus about $0.0061 of output, roughly $0.0107 a day and about $0.32 across 30 days. When a side project is quiet, the arithmetic stays small, and the Relay365 desk keeps the receipt line by line.

Client rules you should read before wiring anything

A third question arrives just before setup: which client will actually work? Relay365 draws firm lines here, and knowing them first saves an afternoon.

Claude models work only from the official Claude Code and Claude Desktop clients, and Claude Code 2.1.285 or newer is required. Cursor, Cherry Studio, Cline, OpenCode, any SDK and cURL pair only with GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax and Hunyuan lanes. Gemini runs on its native protocol only, through Gemini CLI, using :generateContent for single replies and :streamGenerateContent?alt=sse for streams; it cannot go through /v1/chat/completions. Relay365 states these client rules plainly so you can pick the lane that matches your toolchain before you spend anything.

Duty passes for weeks that stop being quiet

Pay-as-you-go covers most side projects, but sometimes a week turns into a sprint. Relay365 offers optional duty passes for Claude and GPT only. A pass comes with a dedicated key that never touches the wallet, and other models always use the wallet.

The Claude pass at 14 days carries $700 of quota with a $380 per week ceiling, and it costs $50.97 for 14 days. You can draw up to $700 of quota at the duty-pass rate x2.5, which is at most about $280 of usage at list prices, and less on models whose pass rate sits above list, with at most $380 of quota per 7 days. The GPT pass at 30 days carries $3000 of quota with a $900 per week ceiling for $81.53 for 30 days, drawing up to $3000 of quota at the duty-pass rate x2.5, at most about $1200 of usage at list prices and less on models whose pass rate sits above list, with at most $900 of quota per 7 days.

Read those numbers as a different shape of commitment, not as a discount. A pass buys headroom for a fixed window, while the prepaid wallet buys freedom to stop. For a project that idles for weeks and then works hard for a weekend, the wallet is usually the simpler instrument; for a sustained stretch of heavy traffic, a duty pass puts the ceiling in writing.

What the desk can trace, refund, or adjust

The last set of questions is about money after the fact. On Relay365, a call that fails upstream with a 502 or 503 is not billed, and a disconnected stream is billed for what was actually used. The dashboard keeps API keys with tags, a key bound to a lane, quick-wiring snippets, and a usage log of the latest 200 requests showing the time, model, input, output, cache recall, charge and result of each call. The log does not show which key made the call, so treat it as a model-level ledger rather than a per-key audit trail.

Top-ups are covered by a straightforward refund rule: within 14 days of a top-up, unused balance can be refunded minus what was consumed and the payment channel's fees. Bounty and promotional credit are not refundable. The referral bounty pays 18% at level 1 and 6% at level 2 of what referred members pay, and it can be withdrawn from $11.11 or moved into the wallet. A new account also receives $0.25 in signup credit, and the account itself is kept in CNY while the rates shown on the pricing page are in approximate USD per 1 million tokens.

One practical limit is worth knowing before you script a retry loop: HTTP 429 applies when there are more than 120 calls in 60 seconds on one key, or more than 8 requests in flight for one account. The desk also keeps a ledger of wallet, bounty and pass accounts, offers in-site messages to the desk once you log in, and presents six interface languages.

Quiet weeks, busy weekends, one transparent table: that is how Relay365 treats a bursty side project. Start with a prepaid wallet, pick the lane that matches your client, and let the usage log tell you where the tokens went. Create a Relay365 account when you are ready to wire it up.