XNews

DeepSeek V4.1 Flash Released with Native Multimodal Support

DeepSeek V4.1 Flash Released with Native Multimodal Support

DeepSeek released V4.1 Flash as a beta model ahead of schedule. It features native multimodal capabilities, a new model architecture, and the same pricing as V4 Flash. The model ID is deepseek-v4.1-flash-expires-on-0910, with a maximum of 20 concurrent requests per account. The release was announced through official WeChat groups and is available for testing. Some users compared it to ZhiPu's GL5.3Flash, expressing less excitement.

Key insights

Release Details and Features

DeepSeek V4.1 Flash is a beta model with native multimodal support, a new model architecture, and pricing identical to V4 Flash. It allows up to 20 concurrent requests per account. The model ID is deepseek-v4.1-flash-expires-on-0910. The release was made early ('超前上线') and is in beta status due to limited resources.

Competitive Landscape and Reception

One user noted that with the availability of ZhiPu's GL5.3Flash, they are less interested in DeepSeek's new model. Another user shared examples showing the model can interpret images (e.g., recognizing pinstripes on a suit). The early beta status suggests DeepSeek is pushing to compete in the multimodal AI space, following OpenAI's Astra which 'proved the value of native multimodal'.

Source posts