DeepSeek-V4.1-Flash is the latest native multimodal model released by DeepSeek. It is a 552B parameter MoE model. It adopts a new Causal-Encoder-Decoder structure, input activation 8B parameters, output activation 16B parameters, and the cost is significantly lower than known models of the same size. After new pre-training and larger-scale reinforcement learning post-training, its benchmark performance exceeds several flagship models including DeepSeek-V4-Pro.
Used in the Tabbit Browser, you can start a conversation directly around the current web page. Whether you are quickly combing long articles, understanding graphic information, or extracting key points, comparing content, and assisting in writing, you can get more timely and fluent answers. The native multimodal capabilities also make it easier to read text and images on web pages, providing efficient help for daily browsing and content processing.
Tabbit Browser is now connected to DeepSeek-V4.1-Flash. Click "Switch Model" at the bottom right of the chat box to switch. You can experience it on both desktop and mobile.