DeepSeek-V4 Flash
by DeepSeek
DeepSeek-V4 Flash is the fast, economical member of the DeepSeek-V4 family, a 284-billion-parameter mixture-of-experts model with 13 billion active parameters and a 1-million-token context window. Its hybrid attention design combines Compressed Sparse Attention and Heavily Compressed Attention, cutting long-context compute and KV-cache size by an order of magnitude versus DeepSeek-V3.2 while scoring 79.0% on SWE-bench Verified.