level="low"Explore
An economical model for well-defined tasks and rapid iterations.
PRICING / DEVELOPER PREVIEWYour search.
Your search.
Your spending limit.
Prepaid credits. A maximum budget per run.
A clear record of what was spent.
EXAMPLE RUN BUDGET$2.00
A user-defined ceiling, not a fixed package price.
result = client.run(
iterations=100,
level="low",
budget=2.0
)Three levels. One interface.
level="middle"Balance
A model level for a balance of inference cost and candidate quality.
level="high"Go deeper
A higher-capacity model for tasks that benefit from more capable generation.
How a run is charged
The estimate endpoint returns conservative operation caps. LLM calls use measured inference cost when available. Sandbox execution and artifacts use operator-configured fixed caps that include service overhead. Unused reserved budget returns to your balance.
When the next step won’t fit
The run ends with BUDGET_EXHAUSTED. You can retrieve its best result. Unknown costs after an interrupted external call are conservatively settled at the reserved cap.
Preview pricing
Model assignments and operation caps are configured by the service operator. Published commercial rates are not yet available. Request an estimate from your API instance before each run.