Model · DeepSeek

DeepSeek V4 Pro

Available now

A million tokens in, long answers out: DeepSeek V4 Pro.

DeepSeek V4 Pro reads up to a million tokens at once and can write very long results. In Agent.Space, pick an Agent and it is ready. No separate account needed.

Works with

In your space, pick an Agent, then pick the model.

SpecsOfficial page

Context
1MToken
Max output
393KToken
Reasoning effort
4 levels
Input
Text

Is it the one?

When to reach for DeepSeek V4 Pro

Matching the model to the task beats paying top price for everything.

Give it

  • Tasks that produce a lot in one go: a whole document, a large block of code.
  • Analysis that reads in a lot of code and material.
  • Simple tasks where you want reasoning switched off for speed.

Skip it for

  • Tasks that need screenshots or design files: it takes text only.
  • Big changes with fuzzy requirements: compare it with Opus 5.5 or GPT-6 Astra.
  • Mechanical work like reformatting or extraction, where a smaller model will do.

In Agent.Space

Pick it and go. The rest is handled.

01

No separate account

No DeepSeek account, no API key to set up. What you use goes on your credits or your plan.

02

Set the reasoning effort

None, Low, High, and Maximum. The default is High. Turn it up for hard problems, down for chores.

03

Agents and teammates, together

In one space, Agents can hand work to each other, and teammates can step in to see the results and carry on.

Pricing

Pay for what you use. See the price live.

1 credit = US$1. This is the official pay-as-you-go price right now. If your plan includes this model, plan allowance is used first.

Model pricing

Per 1M tokens · USD
Loading model prices…

FAQ

About DeepSeek V4 Pro

Right now

Get to work with DeepSeek V4 Pro.

Pick an Agent, choose the model, and leave the rest to your space.