API preview

API preview — not yet available.

Base URL & auth

The API will be OpenAI-compatible. Point any OpenAI SDK or tool at the API host below (available at launch) with your API key. Qwen models served by Alibaba Cloud Model Studio — one key, one balance.

Base URL:  https://app.solver-one.ai/v1   (available at launch)
Header:    Authorization: Bearer YOUR_API_KEY

All keys of an account will share one prepaid balance (USD); a new account starts with a zero balance and is topped up before calling.

curl

curl https://app.solver-one.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen-plus","messages":[{"role":"user","content":"Say hello in Cantonese."}]}'

Python (`pip install openai`)

from openai import OpenAI

client = OpenAI(base_url="https://app.solver-one.ai/v1", api_key="YOUR_API_KEY")

r = client.chat.completions.create(
    model="qwen-plus",
    messages=[{"role": "user", "content": "Say hello in Cantonese."}],
)
print(r.choices[0].message.content)
print(r.usage)   # prompt_tokens / completion_tokens — what you are billed for

Node (`npm install openai`)

import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://app.solver-one.ai/v1", apiKey: "YOUR_API_KEY" });

const r = await client.chat.completions.create({
  model: "qwen-plus",
  messages: [{ role: "user", content: "Say hello in Cantonese." }],
});
console.log(r.choices[0].message.content);

Chatbox / any OpenAI-compatible app

Settings → Model provider: OpenAI API compatible (or "Custom") → API host = https://app.solver-one.ai (some apps want the /v1 suffix, some add it themselves) → API key = your key → model name = qwen-plus (or turbo / max).

Streaming

Pass "stream": true (or stream=True in the SDK) to receive server-sent events (text/event-stream) token by token, ending with data: [DONE].

for chunk in client.chat.completions.create(model="qwen-turbo", messages=[...], stream=True):
    print(chunk.choices[0].delta.content or "", end="", flush=True)

Models

model	best for
`qwen-turbo`	Fastest, lowest cost of the three. Suited to high volume.
`qwen-plus`	Balanced quality and cost.
`qwen-max`	Highest quality for hard tasks.

GET /v1/models (with your key) returns the same list in OpenAI format. Thinking / reasoning mode is off by default on every model (no hidden reasoning tokens on your bill); ask us if you need it.

Error codes

HTTP	`error.code`	message you receive	what to do
400	400	The request was not accepted — check the model name, messages and parameters.	fix the request (model id, `messages`, parameter values)
401	401	Invalid API key, or the key cannot access that model.	missing / invalid / revoked key → check the `Authorization` header; create a new key in your account
403	403	Invalid API key, or the key cannot access that model.	the key cannot use that model → use one of the models listed above
402	402	Insufficient balance — top up to continue.	top up your balance
404	404	The requested model or resource was not found — see GET /v1/models.	use a model id from `GET /v1/models`
422	422	The request body could not be processed — send valid JSON in the chat completions format.	send a valid chat completions JSON body
429	`insufficient_quota`	Insufficient balance — top up to continue.	balance is zero → top up your balance (retrying will not help)
429	`rate_limit_exceeded`	Rate limit exceeded — slow down and retry shortly.	requests / tokens per minute exceeded → back off and retry; honour the `Retry-After` header when present
502	502	The request could not be completed.	retry with backoff; if it persists, contact support
stream	502	`data: {"error": {"message": "The stream could not be completed.", …}}`	the stream ends early; retry the request

Errors are returned as {"error": {"message": "...", "type": "...", "code": ...}} — the shape OpenAI SDKs expect, so they raise the usual exceptions. code is the HTTP status, except on 429, where it tells an empty balance (insufficient_quota) from a rate limit (rate_limit_exceeded) — as OpenAI does.

Limits (preview values; may change at launch)

(larger max_tokens / max_completion_tokens are clamped; n is always 1)

limit	value
requests per minute (per key)	120
tokens per minute (per key)	200,000
max output tokens per request	2,048
balance	prepaid, shared by all keys of the account; calls are refused at a zero balance

Need higher limits? support@solver-one.ai.