DeepSeek V4 Pro
Create an agent or open an existing agent in the builder, then follow the steps below. Model and feature availability may depend on your plan.
What it does
DeepSeek V4 Pro 0813 became selectable, with adjustable reasoning effort.
Before this
An addition to the selector. DeepSeek V4 Flash was already selectable; Pro is the reasoning-heavy sibling — the same million-token window, a far larger output ceiling, and effort levels that start at Fast because thinking is the point of it.
Why it matters
Deep reasoning on hard technical prompts at DeepSeek prices — turn the effort up only where the question deserves it.
The specs
DeepSeek's reasoning-first model, and the cheapest way on the platform to buy a lot of thinking on a hard technical prompt.
- Context
- 1,048,576 tokens in, up to 384,000 out — the largest output ceiling of any text model in the selector.
- Input
- Text only, so it is not the model for an agent that reads screenshots or scans.
- Actions
- Supported.
- Reasoning
- Fast ($$), Smartest ($$$$), or Maximum ($$$$$) — it skips the cheapest tier, because thinking is the point of the model.
- Plans
- Every plan.
The 384k output limit matters for one specific job: generating something long in a single pass — a full report, a migration script, a document rewritten end to end — without chunking it yourself.