GPT-5.6 Sol pay-as-you-go pricing with Relay365 duty desk

From a new key to the first reply at the duty desk

Relay365 is an independent third-party relay run like a round-the-clock duty desk, and the quickest way to understand how it works is to imagine the desk handing you a key and walking you through one request. Every lane sits on one table, so the models you might want are visible in one place instead of scattered across separate tools. The API base is a single address, and the quick-wiring snippets in the dashboard are ready to copy into a project. The pricing page shows the full table of rates, so this article only quotes what helps the worked bill make sense.

Signing up gives a new account $0.25 in signup credit, and the wallet is prepaid: there is no monthly fee, and every call is settled on the token counts the upstream reports. From there the path is short. Create a key, bind the key to the lane you intend to use, paste the matching quick-wiring snippet, and send your first request. The dashboard keeps API keys with tags, so a key that belongs to one lane stays legible to you later.

Two client rules are worth reading before you wire anything. Claude models work only from the official Claude Code and Claude Desktop clients, and Claude Code 2.1.285 or newer is required. The GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax and Hunyuan lanes pair with Cursor, Cherry Studio, Cline, OpenCode, an SDK or cURL. Gemini runs on its native protocol only, through Gemini CLI: a single reply uses :generateContent, a stream uses :streamGenerateContent?alt=sse, and the lane does not go through /v1/chat/completions. Relay365 states these rules plainly because a duty desk that hides its limits is not much of a desk at all.

A worked bill on the Claude Fable 5 lane

Here is an example, not a promise. Suppose you send 2,000,000 input tokens, 500,000 output tokens, 1,000,000 cache write tokens and 2,000,000 cache read tokens for Claude Fable 5 (wire name claude-fable-5) in one day. Relay365 prices the four token types separately, so the bill is built line by line rather than guessed at.

• The input line is 2 million tokens at $3.79 per 1 million tokens, which is $7.58.
• The output line is 0.5 million tokens at $18.94 per 1 million tokens, which is $9.47.
• The cache write line is 1 million tokens at $4.66 per 1 million tokens, which is $4.66.
• The cache read line is 2 million tokens at $0.2917 per 1 million tokens, which is $0.5834.

Adding the four lines: $7.58 + $9.47 + $4.66 + $0.5834 = $22.2934, so this example day comes to about $22.29. If the same volumes repeated for 30 days, the example total would be 30 × $22.2934 = $668.802, or about $668.80. The arithmetic is deliberately boring, and that is the point: you can rebuild it yourself from the published rates.

For comparison, the official list price for Claude Fable 5 output is $50 per 1 million tokens, so the Relay365 output rate of $18.94 is about 38% of official. The official list price for cache read is $1, against $0.2917 here. Relay365 quotes these prices in approximate USD while the account itself is kept in CNY, so the desk shows both views of the same wallet.

The same arithmetic on GPT-5.6 Sol

Now take GPT-5.6 Sol (wire name gpt-5.6-sol) for a second example. Suppose you send 5,000,000 input tokens, 1,000,000 output tokens and 3,000,000 cache read tokens in a day. The input line is 5 × $0.3736 = $1.868. The output line is 1 × $2.09 = $2.09. The cache read line is 3 × $0.0292 = $0.0876. There is no cache write rate listed for this lane, so no cache write line appears on this example bill.

The sum is $1.868 + $2.09 + $0.0876 = $4.0456, or about $4.05 for the example day, and 30 such days would come to 30 × $4.0456 = $121.368, or about $121.37. The official list price for GPT-5.6 Sol output is $20 per 1 million tokens, so the Relay365 output rate of $2.09 is about 10% of official, with the official cache read list price at $0.4 against $0.0292 here.

Two other lanes help show how the table behaves. DeepSeek V4 Flash (wire name deepseek-v4-flash) lists input at $0.2889, output at $1.15 and cache read at $0.006944 per 1 million tokens, with no cache write rate listed. For example, 1,000,000 input tokens, 1,000,000 output tokens and 10,000,000 cache read tokens would total $0.2889 + $1.15 + $0.06944 = $1.50834, or about $1.51. Gemini 2.5 Flash (wire name gemini-2.5-flash) lists input at $0.2389 and output at $1.91 per 1 million tokens, with cache write also at $0.2389 and cache read at $0.0236. No official list price is quoted for these families, so Relay365 makes no comparison for them.

Duty passes, wallet top-ups and the refund line

Relay365 duty passes are optional, and they cover Claude and GPT only; other models always use the wallet. A pass comes with a dedicated key that never touches the wallet, so the two balances stay separate. The Claude pass for 14 days carries $700 of quota at $380 per week, costs $50.97 for 14 days, and draws 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 for 30 days carries $3000 of quota at $900 per week, costs $81.53 for 30 days, and draws at the duty-pass rate x2.5, at most about $1200 of usage at list prices, with at most $900 of quota per 7 days.

Top-ups can be refunded within 14 days: unused balance comes back minus what was consumed and minus the payment channel's fees. Bounty and promotional credit are not refundable, so keep those separate in your head. A failed call is not billed when the upstream returns 502 or 503, and a disconnected stream is billed only for what was actually used. That is the billing rule in full: prepaid wallet, no monthly fee, every call settled on the token counts the upstream reports.

The referral bounty is 18% at level 1 and 6% at level 2 of what referred members pay. It can be withdrawn from $11.11 or moved into the wallet, and the dashboard keeps a ledger of wallet, bounty and pass accounts so you can see which pot a balance came from.

Tracing every request from the desk log

Every request can be traced in the usage log, which is where the duty-desk metaphor becomes concrete. The log holds the latest 200 requests and shows the time, model, input, output, cache recall, charge and result of each call. It does not show which key made the call, so when you are comparing lanes, read the model column and the charge column together with the key tags you set up yourself.

The four token types stay separate all the way through the log, which is why a cache-heavy day and a cache-free day produce visibly different bills. In-site messages to the desk are available once you log in, and the interface is available in six languages. Together with tagged keys, a bound lane, the quick-wiring snippets and the ledger, the dashboard is the place where a Relay365 account is meant to be understood rather than guessed at.

Wiring the last details before you send

Two operational limits are worth keeping in view. If one key sends more than 120 calls in 60 seconds, or if one account has more than 8 requests in flight, the desk will return HTTP 429. Those ceilings are simple to stay under when the quick-wiring snippet is doing the wiring for you, and they keep a single key from becoming the weak point in a busy integration.

So the route from a new key to a first reply is short: sign up, take the $0.25 signup credit, top up the prepaid wallet, create a tagged key, bind it to a lane, copy the quick-wiring snippet, and send. The worked bill above shows what the numbers look like when you add them up honestly, line by line, and the pricing page shows the full table for every lane. Create a Relay365 account when you are ready to put the desk to work on your own traffic.