DeepSeek Flash visual model suddenly updated to support multi-modality. According to news on August 21, DeepSeek has conducted a high-intensity iteration of its large model tool chain and officially upgraded the DeepSeek Harness to 0

Long eyes! DeepSeek Flash visual model suddenly updated to support multi-modality. According to news on August 21, DeepSeek has conducted a high-intensity iteration of its large model tool chain and officially upgraded the DeepSeek Harness to version 0.1.1-rc.1. The biggest highlight of this update is that based on the original DeepSeek V4Flash and V4Pro models, a new visual model experience version-V4Flash Vision-Exp has been added. In the v0.1.0-rc.8 version released the day before, DeepSeek Harness has taken the lead in opening up multi-modal core capabilities. By supporting the configuration of native image requests, the system's key commands such as /goal and /plan officially realize mixed input of images and text. This means that users can now directly send screenshots of their work and specific requirements at the same time, allowing the AI ​​agent (Agent) to directly "look at the pictures and work", which greatly simplifies the development and operation process. Although DeepSeek has previously provided a good image recognition mode on the web page, its deep integration and implementation into the development tool chain marks that V4Flash Vision-Exp has completely completed the last piece of the multi-modal processing puzzle. For developers, just upgrade by specifying the npm command, and you can be the first to experience this new visual model capability that combines speed and accuracy. via AI News (author: AI Base)