Skip to main content
Firecrawl Agent runs on Spark 2, the default model for every run. It removes the accuracy-versus-cost decision the Spark 1 models called for: it is cheaper and faster than both of them while matching Spark 1 Pro’s accuracy, and it completes a higher share of runs.

Available Models

Spark 1 models are deprecated. The spark-1-pro and spark-1-mini names remain accepted for backwards compatibility, but every request — with or without the model parameter — executes on spark-2.

Spark 2

spark-2 is our newest agent model and powers every agent run. Highlights:
  • Lowest cost per run and fastest run time
  • Accuracy comparable to the former Spark 1 flagship
  • Handles everything from simple lookups to multi-domain research
  • The only model with a reasoning budget: pass effort (low, medium, or high) to control how hard it thinks

Spark 1 models (deprecated)

spark-1-pro and spark-1-mini are deprecated. Their names remain accepted for backwards compatibility, but requests that use them route to spark-2 — legacy code keeps working, it just runs Spark 2. There is nothing to migrate.

Specifying a Model

The model parameter is optional — every request runs spark-2:

Pricing by Model

All models use dynamic, credit-based pricing that scales with task complexity:
  • Spark 2: Uses substantially fewer credits than either Spark 1 model for equivalent tasks
  • Spark 1 models (deprecated): Requests route to Spark 2 and are billed accordingly
Credit usage varies based on prompt complexity, data processed, and output structure — regardless of model selected.

API Reference

See the Agent API Reference for complete parameter documentation. Have questions about which model to use? Email help@firecrawl.com.
Are you an AI agent that needs a Firecrawl API key? See firecrawl.dev/agent-onboarding/SKILL.md for automated onboarding instructions.