On September 10, DeepSeek made an official announcement regarding the release of the DeepSeek V4.1 Flash model. Among the new model architecture series, this model stands out as the most compact, boasting native multimodal visual comprehension capabilities. According to tests, the V4.1 Flash surpasses the V4 Pro in multiple aspects, including performance, cost-efficiency, speed, and overall processing time. The structural design of this new model is geared towards achieving higher capability thresholds, faster inference speeds, increased throughput, and the ability to scale up to models with larger parameters.
