GPT-6.1 Sol cURL smoke test on Relay365 API relay

What a cURL smoke test actually proves about a line

At the Relay365 duty desk, a smoke test is the smallest possible handshake with a line: one request, one response, and a clear answer about whether the route is up. Relay365 is an independent third-party relay run like a round-the-clock duty desk, so this check is the first thing we would ask a new caller to run. It is deliberately modest. It tells you that the endpoint accepts your key, that the model identifier resolves, and that tokens come back and get settled. It does not tell you how creative the model is, how it behaves under a heavy prompt, or whether it is the right choice for your product. We would rather say that plainly than dress up a ping as a verdict.

What it does tell you is still worth something. Relay365 keeps every lane on one table, so a check against the GPT-6.1 Sol line runs through the same base address and the same billing account as the rest of your traffic. There is no second dashboard, no separate wallet, and no per-lane onboarding ritual to learn before you can find out whether the route answers.

Every request can also be traced afterwards. The dashboard usage log shows the latest 200 requests with the time, model, input, output, cache recall, charge and result of each call, so a smoke test is not a throwaway ping: it is a line item you can go back and read.

Running the check against the gpt-6.1-sol wire name

Here is the shape of the request. Point cURL at the Relay365 API base shown in your dashboard, use the /v1/chat/completions path, pass your key in the Authorization header, and name the wire model gpt-6.1-sol:

curl $RELAY365_BASE/v1/chat/completions -H "Content-Type: application/json" -H "Authorization: Bearer $RELAY365_API_KEY" -d '{"model":"gpt-6.1-sol","messages":[{"role":"user","content":"Reply with: desk online"}]}'

A healthy result is a normal HTTP 200 with a JSON body, a completion, and a usage block you can compare against the log. If you receive HTTP 429 instead, two known reasons apply: more than 120 calls in 60 seconds on one key, or more than 8 requests in flight for one account. Both are conditions of your own calling pattern rather than evidence that the line is down, so a single 429 is a signal to slow down, not to abandon the lane.

The billing side of a smoke test is simple. Relay365 runs on 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 a 503 is not billed. A disconnected stream is billed for what was actually used, which keeps a truncated response from turning into a surprise charge for work the model never finished.

Reading the four token columns behind the smoke test

Relay365 prices four token types separately: input, output, cache write and cache read. That separation is the reason the usage log lists input, output and cache recall next to the charge, and it is the number you should check after the first call rather than guessing from a single total.

For GPT-6.1 Sol, the rates per 1 million tokens are input $0.1986, output $0.9944 and cache read $0.0389, with no cache write rate listed. The official list price for that model is input $2, output $10 and cache read $0.1, so the Relay365 output rate is about 10% of official. If you want a different GPT line on the same table, GPT-5.6 Sol is priced at input $0.3736, output $2.09 and cache read $0.0292, with an official list price of input $4, output $20 and cache read $0.4, and its Relay365 output rate is also about 10% of official.

One practical note: the rates are quoted in approximate USD while the account itself is kept in CNY, so a dollar estimate and your statement may not match to the last decimal. Treat the figures as a working budget rather than an exact invoice prediction.

The trade-offs we would rather state than hide

An honest relay is not only a list of good numbers, so here is the other side of the desk.

• The usage log shows the latest 200 requests with time, model, input, output, cache recall, charge and result, but it does not show which key made the call, so if you need per-key attribution you should keep your own key tags and records alongside it.
• cURL and SDK clients pair only with the GPT, DeepSeek, Kimi, GLM, Qwen, Grok, MiniMax and Hunyuan lanes, which is exactly why this smoke test targets GPT-6.1 Sol.
• Claude models work only from the official Claude Code and Claude Desktop clients, with Claude Code 2.1.285 or newer, so a cURL smoke test cannot prove a Claude line is up; Claude Fable 5.1 is priced at input $4.00, output $20.01, cache write $4.93 and cache read $0.2375, against an official list price of input $10, output $50 and cache read $0.25, with the Relay365 output rate about 40% of official.
• Gemini runs on its native protocol only, through Gemini CLI, and cannot go through /v1/chat/completions, so Gemini 2.5 Flash at input $0.2389, output $1.91, cache write $0.2389 and cache read $0.0236 is verified in its own client rather than with cURL, and no official list price is quoted for that family.
• Refunds are available within 14 days of a top-up for unused balance, minus what was consumed and the payment channel's fees, while bounty and promotional credit are not refundable.

Duty passes for steady GPT and Claude traffic

If your usage is regular rather than occasional, a duty pass can hold a fixed quota on a dedicated key that never touches the wallet. The GPT pass runs 30 days with $2350 of quota, at most $700 per week, for $81.53. Quota is drawn at the duty-pass rate x2.5, which is at most about $940 of usage at list prices, and less on models whose pass rate sits above list.

The Claude pass follows the same shape: 30 days, $720 of quota, at most $230 per week, for $94.03, again drawing at the duty-pass rate x2.5 for at most about $288 of usage at list prices, and less on models whose pass rate sits above list. Both passes carry their own dedicated key, so the pass quota and the wallet balance stay separate and readable.

Two edge rules are worth knowing before you rely on either card. A call already let through on a pass is drawn in full even if it runs past the quota or the 7-day ceiling, and that overrun counts against the earliest-expiring card. When a pass is nearly used up, each running call holds $3 x the draw rate, and an extra parallel call gets HTTP 429 until the running ones finish. In other words, a pass is predictable in price, but it is not a licence to open unbounded parallel traffic.

Why the desk is still a good deal

Put the rates to work. For example, if you send 1,000,000 input tokens and 100,000 output tokens a day on GPT-6.1 Sol, that is $0.1986 plus $0.09944, or about $0.30 a day and about $8.94 over 30 days. Add cache reads at $0.0389 per 1 million tokens when your prompts reuse context, and the ledger keeps every one of those four token types visible rather than collapsing them into a single opaque line.

Relay365 also keeps the small things simple. A new account receives $0.25 in signup credit. The referral bounty pays 18% at level 1 and 6% at level 2 of what referred members pay, withdrawable from $11.11 or moved into the wallet. The dashboard holds API keys with tags, a key bound to a lane, quick-wiring snippets, a ledger of wallet, bounty and pass accounts, in-site messages to the desk, and six interface languages.

So the honest summary is this: a cURL smoke test proves a line is up and priced as advertised, but it cannot tell you everything about a model, and the usage log will not tell you which key made a call. What you do get is one table of lanes, separately priced token types, a traceable request history, and GPT output at about 10% of official list. That is a workable balance, and it is why the desk keeps the lights on.

Create a Relay365 account, run one smoke test against the GPT-6.1 Sol line, and read the result in your own usage log before you commit real traffic.