Qwen3.8 Max cURL smoke test on the Relay365 API relay

A note from the night desk about the spend you keep losing

Hello from the Relay365 duty desk. It is late, the lamps are on, and someone on our side is always awake watching the lanes. You wrote to us because your AI spend keeps slipping away from you. One week it looks small, the next week the balance has moved and you cannot say which call moved it. That is a bookkeeping problem before it is a pricing problem, and it is exactly the problem Relay365 is built to answer.

Relay365 is an independent third-party relay, run like a round-the-clock duty desk. Every lane sits on one table here, so you are not juggling a dozen accounts just to see what you spend. Every request can be traced in the usage log, and the four token types are priced separately: input, output, cache write and cache read. When those four are broken out, you can look at a single line in the log and say what it cost and why.

We want to hand you one small ritual tonight: a cURL smoke test that proves a Qwen3.8 Max line is up before you lean on it. It is a short command, one response, and then a glance at the Relay365 dashboard to see the charge and the result recorded.

What a Qwen3.8 Max smoke test actually checks

The wire name you want is qwen3.8-max. Point your cURL request at the API base for Relay365 and ask for one short reply. That single call settles three questions at once. It shows the lane answers. It shows the token counts the upstream reports for your call. And it leaves a row in the usage log with the time, the model, the input, the output, the cache recall, the charge and the result.

Keep the test small on purpose. A handful of input tokens and a one-line answer is enough to see the lane respond and to see the four token types priced on their own. On Relay365, Qwen3.8 Max is listed at input $0.8028, output $2.41 and cache read $0.0958 per 1 million tokens, with no cache write rate quoted for this lane. Rates are per 1 million tokens in approximate USD, and the account itself is kept in CNY.

One habit we suggest while you are here: tag the key you use for tests. The Relay365 dashboard lets you create API keys with tags and even bind a key to a lane, so the smoke-test key stays obvious. The usage log shows the latest 200 requests and does not show which key made a call, so the tags are how you keep your own meaning straight.

Reading the four token types on one row

Because input, output, cache write and cache read are priced separately, a Qwen3.8 Max row tells you more than a single blended number ever would. Say you send 200,000 input tokens a day and the lane returns 50,000 output tokens. On the Relay365 Qwen3.8 Max rates that is, for example, about $0.16 of input and about $0.12 of output per day, roughly $0.28 before any cache read. That is an example worked only from the rates on our table, and it is the kind of sum you can now do yourself from your own log.

Cache read deserves its own glance. On this lane it is $0.0958 per 1 million tokens, and it is billed at its own rate rather than folded into input. If your calls lean on cached context, that is the line that will move, and you will see it move on its own.

Billing on Relay365 is a prepaid wallet with no monthly fee, and every call is settled on the token counts the upstream reports. A call that fails upstream with a 502 or 503 is not billed. A disconnected stream is billed for what was actually used. So the smoke test is cheap to run and honest about what it costs.

Where the official prices come in, and where they do not

For Qwen3.8 Max we quote no official list price for this family, so we make no comparison and neither should you. Where we can compare, we do. On the Claude lanes, Relay365 lists Claude Fable 5 at input $3.79 and output $18.94 per 1 million tokens, against official list prices of $10 and $50, so the Relay365 output rate is about 38% of official. On the GPT lanes, GPT-6 Astra is listed at input $0.7444 and output $3.72 against official list prices of $10 and $50, so the Relay365 output rate is about 7% of official.

Keep the client rules in view as you plan. Claude models run only from the official Claude Code and Claude Desktop clients, Claude Code 2.1.285 or newer, and not from cURL. Gemini runs on its native protocol only, through Gemini CLI. The clients such as Cursor, Cherry Studio, Cline and OpenCode pair with the GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax and Hunyuan lanes, which is why a cURL smoke test is the right instrument for your Qwen3.8 Max line.

The desk habits that keep the number visible

Run the smoke test whenever you open a new key or after a quiet period, and read the row it writes. Then keep the wallet honest. A new account starts with a signup credit of $0.25, refunds of unused balance are possible within 14 days of a top-up minus what was consumed and the payment channel's fees, and bounty and promotional credit are not refundable. There is also a referral bounty of 18% at level 1 and 6% at level 2 of what referred members pay, withdrawable from $11.11 or movable into the wallet.

If your Claude work grows heavy, there are duty passes for Claude and GPT only, each with a dedicated key that never touches the wallet. The Claude 14-day pass is $50.97 and draws $700 of official-price quota at official price x2.5, about $280 of official-priced usage, with at most $380 of quota per 7 days. The GPT 30-day pass at $3000 is $81.53 and draws $3000 of official-price quota at official price x2.5, about $1200 of official-priced usage, with at most $900 of quota per 7 days. Everything else, including your Qwen3.8 Max line, always runs on the wallet.

That is the whole trick: one table, one log, four token prices, and a one-command test you can run at any hour. Create a Relay365 account and try the Qwen3.8 Max smoke test tonight.