Qwen3.8 API pricing
Three monthly quotas, one fixed model, no unlimited-plan fiction. Taxes are calculated at checkout, and you can cancel future renewals from your billing portal.
Solo builders
$12 / month
A practical first step for prototypes and personal tools.
Product teams
$49 / month
More room for production workflows and growing products.
Growing businesses
$189 / month
Predictable capacity for higher-volume applications.
Cost comparison
These subscriptions are call-based, not token-priced. The effective figure below divides the monthly price by the included call quota; it is a planning number, not a separate metered charge. Unused calls do not roll into the next UTC month.
| Plan | Monthly price | Included calls | Price per included call | Best planning fit |
|---|---|---|---|---|
| Starter | $12 | 300 | $0.040 | Solo builders |
| Pro | $49 | 1,500 | $0.033 | Product teams |
| Scale | $189 | 6,000 | $0.032 | Growing businesses |
Taxes may be added at checkout. A call accepts up to 7,500 UTF-8 bytes of combined message content and can request up to 1,024 output tokens, but actual usage varies by prompt and response. The quota is a maximum number of accepted calls, not a promise that every task needs the same compute.
Over a 30-day month, Starter averages 10 included calls per day, Pro 50, and Scale 200. Real traffic is rarely even, so reserve room for launch spikes, staging, support investigations, and a small number of bounded retries. If one customer action triggers several completions, budget for the whole workflow rather than the visible button click.
Track accepted calls per product action for a week before moving up a tier. The cheapest plan is the smallest one that covers measured demand with safe headroom—not automatically the lowest monthly sticker price.
Billing details
What the price includes
Every tier uses the same fixed Qwen3.8 27B model route, request validation, API key management, monthly usage counter, and non-streaming response shape. Higher tiers buy more monthly calls; they do not unlock a different model.
No. This is an independent call-based service with its own limits. The official model and managed Qwen services have separate owners, terms, features, and prices.
The Worker reserves a call before contacting the provider. When it can confirm a network or provider failure, it releases that reservation.
No. This endpoint currently accepts text messages and returns one completed response. Use another service if images, tools, streaming, or custom routing are required.
Estimate calls from a real week, include bounded retries and staging traffic, then choose the smallest tier with headroom. Upgrade only when measured use justifies it.
Free option
Use the homepage Space to explore prompts before buying. It is a third-party community service, so capacity, uptime, and data handling are separate from the paid first-party API.