Live model prices are available in the console
Model and media billing
Operating subscriptions and usage charges are separate. Standalone use of CLI, serve, ACP or headless does not require an SDK subscription; platform model and media calls are still billed by usage.
Choose an operating scale for your applications
SDK plan prices, free application limits and operating benefits come from the live catalog. Services delivered through platform applications use the limits of their individual or organization owner.
View SDK pricing →Individuals · usage-based billing
Check available models, current prices and your wallet balance in the console.
- For platform-provided models, the platform pays the provider and settles model usage with you.
- Input, output and supported cache usage are recorded separately, with rates listed for each model.
- Usage records include request identifiers to help reconcile individual calls with your bill.
- Direct calls to local or custom endpoints are tracked separately from platform-paid calls.
Organizations · shared ownership and billing
Choose the correct individual or organization billing owner.
- Organization applications use that organization's plan, operating limits and wallet.
- Applications remain owned by the organization when their creator leaves; deletion requires organization consent.
- Model availability, member permissions and governance settings together control access.
- Negotiated rates, support and dedicated deployment follow the actual agreement and console quote.
What is metered
These are metering principles, not a fixed price list. Check the console catalog and actual bill for each model's units, dimensions, specification conditions and rates.
- LLM · input and output
- Input and output tokens are billed separately, usually quoted per million tokens. Provider usage receipts take precedence; missing usage follows the estimation and labeling rules of the relevant request path.
- LLM · cache reads and writes
- Models that support cache pricing list separate cache dimensions. Their active billing rules determine how cached tokens are accounted for without also charging them as ordinary input.
- Image generation and editing
- Depending on the model, usage is metered by images, calls or token dimensions. Resolution, quality and other specifications may carry different rates.
- Video generation
- Depending on the model, usage is metered by duration, calls or token dimensions. The catalog specifies the applicable configuration and billing unit.
- Speech recognition (ASR) and speech synthesis (TTS)
- ASR may be metered by audio duration or tokens; TTS may be metered by characters. Units are listed separately for each model.
- Web search
- Charges follow the selected search service's billing items and usage. Search is not covered by bonus LLM token allowances.
- Local direct calls and bring your own key (BYOK)
- Direct requests to local or custom endpoints are not platform-paid model usage. Calls using your provider key follow the provider's or your contract's billing rules. No platform model deduction does not mean the provider is free.
Billing and usage details
How do subscriptions differ from model charges?
SDK subscriptions provide application operating capacity. Model and media calls are charged separately by usage. Free application creation does not mean free model usage. Choose an individual or organization before purchasing; the server quote determines the amount due, benefits and effective dates.
How is concurrency counted?
Under the same billing owner, sessions active in the last five minutes share concurrency limits according to the runtime rules. Direct server calls without a session ID are counted as in-flight requests: each occupies one slot, with a two-minute slot expiry. The hard cap rejects new reservations; sessions returning after the heartbeat window are checked again. Individual media and search calls do not occupy concurrency slots.
What can bonus tokens offset?
Only platform-paid LLM usage settled in Chinese yuan (CNY). Images, video, ASR, TTS, search, BYOK and client-reported usage are excluded. The purchase order determines the allowance, effective date and expiry limits. Bonus tokens cannot be redeemed for cash.
Which payment and top-up methods are supported?
Use the payment or redemption options currently available in the console. Availability can vary by account, region or service configuration. Check the amount and billing owner before paying.
What happens if my balance is insufficient?
The platform checks available funds for calls that require prepayment; insufficient funds may cause new requests to be rejected. Check your balance, existing reservations and top-up records in the console before retrying.
Do application and model capabilities change by tier?
The SDK's adjudication, memory, context, tool and media mechanisms span the plans. Actual availability also depends on application configuration, permissions and model support. Operating limits, gifts and agreed services follow the live catalog or valid contract.
Check the rates, then start building
Find the current model catalog, plan quotes and account usage in the console.