Plans determine how cheaply and how deeply you can run it.
Subscription = access + priority + discount. Usage = metered execution.
You pay for access to an execution engine. Your plan determines how cheaply you can run jobs on it. Every job shows an estimate before you run. No surprise bills.
Row counts are guidance. Actual usage varies by job complexity (model + web depth).
Subscription = permission + priority + discount. You get access to bulk.run's execution engine, priority in the queue, and a discounted execution rate. Usage is then metered based on what your jobs actually do — AI calls, web search, tools.
A row is not a unit of work — it's a unit of addressing. Classifying a sentence costs differently than deep web research. Execution-based pricing is honest: you pay for what the job actually costs, and you see the estimate upfront.
New users get a one-time $3 execution grant to experience bulk.run. It's not a free plan — execution isn't free. We just want you to try it once before committing.
Every job shows an estimated cost range (e.g., "$1.40–$1.80") before you run it. The estimate depends on row count, model choice, and web search depth. No surprises.
All paid plans include access to Google Gemini 2.5 models (Fast, Balanced, Max). Faster models cost less per job. You choose the tradeoff.