The founder of DeepSeek announced the release date of the flagship model V4

DeepSeek is preparing V4: release date and new features

*The founder of DeepSeek – Liang Wenfeng – confirmed that the next-generation flagship model, DeepSeek V4, will be officially launched at the end of April 2026. The information came from internal communication with employees and has been verified by AIBase.*

What’s new in V4?
ModePurposeFeaturesFast Mode (quick mode) Everyday conversations, instant responses Support for text recognition in images and files; focused on speed Expert Mode (expert mode) Solving complex logic tasks and deep analysis Enhanced intelligent search. File uploads and multimodal features are not yet supported; queues may occur during peak hours
*The important innovation is a multi-level operating mode, first introduced in DeepSeek.*

Current situation
- Positive feedback: users note significant progress in logical data processing and programming functionality.

- Technical glitches: over the past three days the platform has experienced major issues, including one incident lasting up to 12 hours. Experts call it a “painful transition period” from the old model to the new one.

Competition
*DeepSeek chose its release date amid intense competition:*

- Tencent – Yao Shunyu’s team plans to launch Hunyuan next month.

- This means that the end of April will become a period of direct rivalry between two leading Chinese developers of foundational AI models, which could shift the balance of power in the industry.