Billing and limits

Prepaid credits#

Seldon bills from prepaid workspace credits, in USD. Add credits under Add credits in Billing by card checkout through Stripe, or redeem a promo code. Billing also lists recent ledger entries.

When a workspace has no available credit, gateway requests return 402 PAYMENT_REQUIRED, except responses served from the cache, which are free.

What a gateway request costs#

Each non-streaming response reports usage.computed_cost: the tokens used, priced at the serving offering's rates, including any discounted cached input tokens. Streamed responses are priced the same way; their cost appears in Calls & usage. Model prices are listed on Models.

When structured-output enforcement asks a provider again, every attempt is counted.

What Functions cost#

Functions have two separate kinds of spend:

Compilation charges actual usage even when no executable alternative is found or the run is cancelled. While a provider call's outcome is uncertain, its estimated cost stays held rather than released; the run summary shows spent and held amounts separately.

Each function call reserves up to $10 while it runs (revisions created before this limit use $1) and is charged for the tokens of every stage that ran. Functions always run on workspace credits, not BYOK.

Spend and rate limits#

• Workspace balance. Gateway requests and function calls stop when credits run out.
• API key limits. When you create a key, Advanced offers a spend limit and a requests-per-minute limit. They currently apply to function calls.
• BYOK budgets limit spend on your own provider credentials. BYOK usage never counts against workspace spend limits. See Bring your own key.

Limits#

See Limits for every numeric limit.