DeepSeek V4 (Pro / Flash)
Open-weight (MIT-licensed) model family that shipped April 24, 2026. V4-Pro is a 1.6T-parameter MoE model (49B active) for complex reasoning/coding/agentic work; V4-Flash is a lighter 284B-parameter model (13B active) for speed and cost.
Source · verified
DeepSeek V4 in the news
DeepSeek open-sources Super-Agent, a coding agent that watches team chats and opens pull requestsDHS orders DeepSeek products, and third-party products built on them, out of federal civilian agenciesDeepSeek ships V4.1 Flash, a 552-billion-parameter model built around a split encoder-decoder designA single attacker used hundreds of AI agents to breach 395 organizations in hoursAI startup Magic claims a pretraining recipe roughly 50x more compute-efficient than DeepSeek's