Introducing Exa Deep: putting an agent inside every search For each query, an agent runs in a loop until it gathers all information, then returns structured output. Evals show Deep is Pareto optimal at 4-60s latency, ideal for quick, cost-efficient research!

We optimized Exa Deep for cost and speed. Under the hood, an agent decomposes every search into a plan with multiple rounds of parallel sub-searches. Sub-searches use our <200ms Exa Instant endpoint to reduce latency.

Deep is highly controllable. It can handle complex input queries with complex instructions. You can define an output schema, and Deep returns structured JSON with field-level citations: