GPT-5.6 Sol pricing on Relay365: your duty desk walkthrough

Welcome to the Relay365 duty desk. The duty desk keeps watch over an AI API relay run like a round-the-clock shift: every lane is lined up on one table, every request can be traced, and every token type is priced on its own line. This walk follows the pages a new member opens, in the order they open them, and ends with a plain break-even sum for GPT-5.6 Sol. Relay365 is an independent third-party relay, so the prices you read are Relay365 prices, not a pass-through of any vendor's own billing.

The pricing page: one table for every lane

You arrive at the pricing page before you create anything, and the layout tells you how the desk works. Every lane sits on one table, with rates per 1 million tokens in approximate USD, while the account itself is kept in CNY. The table separates four token types and prices each of them: input, output, cache write and cache read. That separation matters, because a workload heavy on cached context and a workload heavy on fresh output do not cost the same, and Relay365 shows both sides of that picture instead of hiding them in one blended number.

Look at GPT-5.6 Sol, whose wire name is gpt-5.6-sol. Input is $0.3736 per 1 million tokens, output is $2.09, cache read is $0.0292, and there is no cache write rate for this lane. Against the official list price of $4 input, $20 output and $0.4 cache read, the Relay365 output rate is about 10% of official. The same page also lists GPT-6.1 Sol at $0.1986 input, $0.9944 output and $0.0389 cache read, with its output rate about 10% of official. Claude Fable 5 sits on the same table too, at $3.79 input, $18.94 output, $4.66 cache write and $0.2917 cache read, with an output rate about 38% of official. One page, one comparison, no hunting.

Two other columns deserve attention before you scroll on. The GPT and Claude columns carry a duty pass option, the desk's version of a 值守卡, while other lanes always draw from the wallet. And the desk keeps a note on client rules right where you make your choice: Claude models work only from the official Claude Code and Claude Desktop clients, with Claude Code 2.1.285 or newer, while GPT lanes pair with Cursor, Cherry Studio, Cline, OpenCode, SDKs and cURL. Gemini has its own native protocol through Gemini CLI. Knowing this on the pricing page saves an afternoon later.

The signup page and the $0.25 welcome credit

Next comes the signup page. A new Relay365 account receives $0.25 in signup credit, which is small enough to be a nudge rather than a promise, but enough to place a small request and see the shape of the system. The billing model is prepaid: there is a wallet, there is no monthly fee, and every call is settled on the token counts the upstream reports. Nothing is billed on a timer or on a subscription basis.

The desk also spells out the settlement edges here, because a relay should be readable before it is used. A call that fails upstream with a 502 or 503 is not billed at all. A disconnected stream is billed for what was actually used, since those tokens were genuinely delivered. Refunds are available within 14 days of a top-up, with unused balance returned minus what was consumed and the payment channel's fees, while bounty and promotional credit are not refundable. Referral bounty runs at 18% at level 1 and 6% at level 2 of what referred members pay, withdrawable from $11.11 or moved into the wallet.

There is a second path on this page for members who expect a heavy month. Instead of topping up the wallet, you can take a duty pass, and the pass arrives with a dedicated key that never touches the wallet. That separation is the reason a pass can be compared to wallet spending directly: the money is committed up front, and the key draws from the pass.

The dashboard: keys, lanes and quick-wiring snippets

Once you are inside the dashboard, the desk hands you the controls. You create API keys, tag them, and bind each key to a lane, so a key that talks to GPT-5.6 Sol is a GPT key and nothing else quietly shares it. Tags make sense of a busy project later, when several services call the same relay for different purposes.

Beside the key list sit quick-wiring snippets. These are the short lines you paste into your own application, and they are where the client rules become practical. A GPT lane can be wired into Cursor, an SDK or a cURL call. A Claude lane needs the official Claude Code or Claude Desktop client, and it will not run from those other tools. Gemini travels on its native protocol through Gemini CLI, using :generateContent for a single reply and :streamGenerateContent?alt=sse for a stream, and it does not go through /v1/chat/completions. The snippets are there so you wire the lane the way it was designed to be wired.

The dashboard also shows the desk's traffic rules in plain view. More than 120 calls in 60 seconds on one key returns HTTP 429, as does more than 8 requests in flight for one account. Six interface languages are available, so the desk can be read in the language you work in. And when a question comes up, in-site messages go straight to the desk, once you are logged in.

The duty pass page: reading the GPT card

The duty pass page is where the numbers on the pricing table become a commitment. Two GPT cards are listed. The first is 30 days, $2350 of quota, $700 per week, priced at $81.53. The second is 30 days, $5650 of quota, $1640 per week, priced at $194.03. Both draw quota at the duty-pass rate x2.5, which means up to $2350 of quota covers at most about $940 of usage at list prices, and up to $5650 covers at most about $2260, less on models whose pass rate sits above list. The weekly ceiling applies per 7 days.

Two desk notes make the card behave predictably. A call already let through on a pass is drawn in full from the pass even if it runs past the quota or the 7-day ceiling, and that overrun counts against the earliest-expiring card. And when a pass is nearly used up, each running call holds $3 x the draw rate, while an extra parallel call receives HTTP 429 until the running ones finish. In plain terms, a pass behaves like a reserved shift slot: work already on the desk finishes, and new parallel work waits rather than piling on.

The break-even walk for GPT-5.6 Sol

Now the sum. On the wallet, GPT-5.6 Sol costs $0.3736 per 1 million input tokens, $2.09 per 1 million output tokens and $0.0292 per 1 million cache read tokens. For example, if you send 1,000,000 input tokens, 1,108,000 output tokens and 1,000,000 cache read tokens a day for 30 days, the wallet bill is about $81.56. The $81.53 GPT pass for 30 days therefore lands right at your break-even point for that pattern of use, and for heavier output months the pass moves ahead.

Run the same volume against the official list price and the contrast is clear. Input at $4, output at $20 and cache read at $0.4 per 1 million tokens puts that identical 30-day pattern at $796.80. The pass covers up to about $940 of usage at list prices, so this workload fits inside the card with room to spare, and the wallet rate would have cost roughly $81.56 against a pass price of $81.53.

Which one fits depends on the shape of your month. If usage is steady and predictable, the wallet keeps the flexibility, with no monthly fee and settlement on reported token counts. If you know a heavy 30-day stretch is coming and would rather commit the money up front, the pass fixes the cost at $81.53 for up to $2350 of quota at $700 per week, and its dedicated key keeps the wallet untouched.

The usage log, the ledger and the desk

The last pages a new member opens are the ones that make the earlier choices accountable. The usage log shows the latest 200 requests with the time, model, input, output, cache recall, charge and result of each call, so every request can be traced back to a line, even though it does not show which key made the call. The ledger keeps the wallet, bounty and pass accounts side by side, so you can see which pot a draw came from.

That is the full round on Relay365: one pricing table for every lane, four separately priced token types, a wallet with no monthly fee, optional duty passes for Claude and GPT, and a duty desk that keeps the log open behind every call. When you are ready to start your own round, create a Relay365 account and open the pricing table for yourself.