DeepSeek has released its newest AI fashions, the V4 Professional and Flash variations, a bit over a 12 months after it went viral and became the top rated free app on Apple’s App Retailer within the US. “Welcome to the period of cost-effective 1 million context size,” DeepSeek mentioned in its announcement. Context size is what you name the utmost variety of tokens that an AI mannequin can bear in mind, so the larger it’s, the extra coherent and constant an AI is in the case of prolonged conversations. OpenAI’s recently announced GPT‑5.5 has a context window starting from 400,000 to 1 million, as an illustration.
The brand new mannequin remains to be open-source, permitting customers to obtain its code and modify it if they need. DeepSeek says V4 Professional has enhanced agentic capabilities and claims that it rivals high closed-source fashions in the case of reasoning. It additionally says that it trails solely Gemini-3.1-Professional in wealthy world information. In the meantime, V4 Flash isn’t fairly as highly effective because the V4 Professional, but it surely has sooner response instances. Nonetheless, its reasoning skills carefully method V4 Professional, DeepSeek says, and it performs on par with with the Professional model on easy Agent duties.
Shortly after DeepSeek topped the App Retailer charts, it was banned to be used by US federal businesses and on government-owned units. Authorities believed it was a nationwide safety danger and posed a menace to US AI shares. South Korea additionally paused downloads of its app over privateness issues.
