What is DeepSeek V4 Pro?
DeepSeek V4 Pro is the open-weight flagship of the DeepSeek-V4 series, released in general availability on August 13, 2026. It is a mixture-of-experts model with 1.6T total parameters and 49B active per token, and it handles contexts up to 1,000,000 tokens.
Key Features
- 1.6T parameters with 49B activated per token, published under the MIT license
- 1,000,000 token context window using hybrid compressed attention
- Reasoning effort levels of low, high and max for agent workflows
- Native OpenAI Responses API support, with a one-click Codex setup
- Peak and off-peak API pricing, where off-peak rates are 50% lower
How to Access
Use the DeepSeek platform API, the official chat app, or OpenRouter. Weights and the model card are published on Hugging Face, so you can self-host with transformers and a suitable inference engine.
