DeepSeek V4 Pro

DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: (1) a hybrid attention architecture that combines Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) to improve long-context efficiency; (2) ManifoldConstrained Hyper-Connections (mHC) that enhance conventional residual connections; (3) and the Muon optimizer for faster convergence and greater training stability

Model overview

Model ID: deepseek/deepseek-v4-pro
Provider: deepseek
Type: language
Context window: 1,000,000 tokens
Max output tokens: 384,000

Tags: implicit-caching, reasoning, tool-use

Model pricing

Metric	Value
`Input tokens (/1M)`	`$0.66`
`Output tokens (/1M)`	`$1.98`
`Image generation`	`n/a`
`Cached input read (/1M)`	`$0.02`
`Cached input write (/1M)`	`n/a`
`Pricing source`	`gateway`

Related docs

Model catalog

Browse every generated model page.

Open catalog

Pricing matrix

Compare model pricing across the full catalog.

Open pricing matrix