Chinese AI lab DeepSeek has released V4, its first major model update in over a year
DeepSeek has released DeepSeek V4 Pro and DeepSeek V4 Flash, its first major model update in over a year. The new V4 models introduce a context window of up to 1 million tokens, allowing longer conversations and documents to stay more coherent at a lower cost.
DeepSeek V4 remains open-source, allowing users to inspect, adapt, or modify the code. V4 Pro is aimed at more complex, multi-step agentic tasks, with DeepSeek claiming it matches leading closed-source models in reasoning and trails only Google’s Gemini 3.1 Pro in world knowledge. V4 Flash is the faster variant, offering quicker responses while staying close to V4 Pro on simpler agent tasks.
The release also comes amid growing scrutiny. DeepSeek has been banned for use by US federal agencies and on government devices over national security concerns, while South Korea temporarily suspended downloads over privacy concerns before later restoring availability.
- AI Chatbot
- Freemium
- Open Source
Related news
External links
- DeepSeek-V4: Towards Highly Efficient Million-Token Context IntelligenceDeepSeek on Hugging Face • Official source
- DeepSeek-V4: a million-token context that agents can actually useHugging Face Blog • Official source
- China's DeepSeek launches an update of its AI modelAP News
- China’s AI upstart DeepSeek drops new model. Will it make waves like last year?CNN
- DeepSeek promises its new AI model has 'world-class' reasoningEngadget
- China's DeepSeek releases preview of long-awaited V4 model as AI race intensifiesCNBC
Comments
The flash deepseek v4 is super good and way cheaper than v4 pro via the api so unless your tasks require that extra bit of reasoning its not worth paying multiple times more per token. just an fyi