DeepSeek V4 Pro on Relay365: an API relay walk through

Opening the Relay365 table after sign-up

Welcome to the Relay365 duty desk. You have just created an account, and the first thing in front of you is not a contract but a table: every lane Relay365 supports sits on the same pricing page, with its wire name, its token rates, and the four token types priced separately. New accounts also receive $0.25 in signup credit, which you can put toward a first call while you learn the interface.

Relay365 is an independent third-party relay run like a round-the-clock duty desk. The account itself is kept in CNY, while the rates on the table are shown per 1 million tokens in approximate USD, so you can read them next to the pricing you already know. Billing runs on a prepaid wallet with no monthly fee, and every call is settled on the token counts the upstream reports. Nothing is billed for a call that fails upstream with a 502 or 503, and a disconnected stream is billed only for what was actually used.

From here, a walk through the pages in the order you will open them: the pricing table first, then the key page where you bind a key to a lane, then the usage log, then the ledger that shows your wallet, bounty and pass accounts. The whole point of that order is simple. You pick a model on one table, wire it once, and then watch every request land in a log you can read.

Reading the four token columns for DeepSeek V4 Pro

Start with DeepSeek V4 Pro, whose wire name is deepseek-v4-pro. On the Relay365 table it is listed at $1.45 for input and $4.35 for output per 1 million tokens, with cache read at $0.0486 and no cache write rate listed. Because input, output, cache write and cache read are priced separately, you can see exactly which part of a request costs what, and you can compare a cached run with a fresh one without guessing.

Here is a plain example: if you send 1 million input tokens and 100,000 output tokens a day to DeepSeek V4 Pro, the input side is $1.45 and the output side is $0.435, for $1.885 a day and $56.55 over 30 days. Add 1 million cache read tokens on top and you add $0.0486. No official list price is quoted for this family, so the Relay365 desk makes no comparison for it; the numbers on the table are simply the numbers you pay.

Where a comparison does exist, the desk will point it out. Claude Fable 5 carries an output rate of about 38% of the official list price, and GPT-6 Astra an output rate of about 7%. Both sit on the same table as DeepSeek V4 Pro, which is the point of a single relay: comparing a Claude lane, a GPT lane and a DeepSeek lane is a matter of reading across one page rather than opening three separate dashboards.

Binding a key to a lane and wiring it in

The second page you open is the key page. Relay365 lets you create API keys with tags, and you bind a key to a lane, so a key that is meant for DeepSeek stays pointed at DeepSeek. Quick-wiring snippets sit next to the key list, so you can drop the base endpoint into your project and send a first request without assembling the request by hand.

Client rules are firm, and the desk would rather state them plainly than let you find them later. Claude models work only from the official Claude Code and Claude Desktop clients, with Claude Code 2.1.285 or newer. Cursor, Cherry Studio, Cline, OpenCode, any SDK and cURL pair only with the 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, and it cannot go through /v1/chat/completions.

So if your plan is to try DeepSeek V4 Pro while another model stays in your stack, you can keep both on Relay365 without opening a new subscription. The DeepSeek lane and the GPT lane sit side by side on the same table, and the key you tag for one does not need to be confused with the key you tag for the other.

Tracing every call in the usage log

Third page: the usage log. It shows the latest 200 requests, and for each one the time, the model, input, output, cache recall, the charge and the result. One thing the desk will tell you plainly is that the log does not show which key made the call, so if you need to attribute a request, use key tags and lanes for that. The log is the place where you check a DeepSeek V4 Pro trial against the rate you expected to see.

Beside it sits the ledger of wallet, bounty and pass accounts, so you can see what is available on the prepaid wallet and what sits in a duty pass without moving money around. There are also in-site messages to the desk, available once you log in, six interface languages, and a limit structure worth knowing on day one: HTTP 429 appears when a single key sends more than 120 calls in 60 seconds, or when one account keeps more than 8 requests in flight.

Because every request can be traced in the usage log, a trial becomes a measurement rather than a feeling. You can see how many output tokens a prompt actually produced, how much cache recall came back, and what the call was charged, then decide whether the lane deserves a place in your default rotation.

Choosing a duty pass and closing the shift

If you want to try a model without a new subscription, the duty passes on Relay365 are the last page to look at, and they are optional. A Claude pass of 14 days costs $50.97 and draws up to $390 of quota at the duty-pass rate x2.5, with at most $210 of quota per 7 days. A GPT pass of 30 days costs $81.53 and draws up to $2350 of quota at the duty-pass rate x2.5, with at most $700 of quota per 7 days. Each pass comes with a dedicated key that never touches the wallet, and other models always use the wallet.

Two rules are worth carrying to your next call. 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 the 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 gets HTTP 429 until the running ones finish.

Money matters here too. 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.

That is the whole shift at the Relay365 duty desk: one table of lanes, one key per lane, one log that traces every request, and a wallet that settles on the tokens the upstream reports. When you are ready to take the seat, create a Relay365 account and open the table yourself.