Search
On September 10, DeepSeek officially introduced the V4.1 Flash model, the most compact iteration within its innovative structural lineup. This model boasts native multimodal visual comprehension abilities. Leveraging a Causal-Encoder-Decoder architecture with asymmetric input-output configurations, it dramatically slashes inference expenses. Furthermore, the V4.1 Flash outperforms its predecessor, the V4 Pro, in every aspect, encompassing performance, cost-efficiency, and processing speed. Consequently, DeepSeek has outlined a strategic plan to gradually discontinue the V4 Pro model. The supercomputing internet swiftly rolled out API services and weight files tailored for the V4.1 Flash. Moreover, the V4.1 Flash adopts a peak-valley pricing strategy, with the updated pricing structure commencing at 12:00 PM on September 10.
-
C114 Communication Network -
Communication Home
