TLDR Tech / AI
DeepSeek-V4.1-Flash
DeepSeek-V4.1-Flash is designed for greater capability, faster inference, higher throughput, and scaling to larger models. It uses an asymmetric architecture that gives it more intelligence at less cost. The model's smaller KV cache provides users with bigger savings. The model is now live on the DeepSeek API with native multimodal support. V4-Flash and V4-Flash-Vision-Exp have been retired.
Read the full story