DS
DeepSeekDeepSeek-V4-Flash
V4 high-speed, low-cost edition — strong reasoning, very low call cost
Parameters284B total (13B active per inference)Context Window1M tokens (max output 384K)Knowledge CutoffNot publicly disclosed
Overview
DeepSeek-V4-Flash is the high-speed, low-cost edition in the V4 flagship line, pairing strong reasoning with very low invocation cost for high-frequency production use.
It complements V4-Pro within the latest V4 generation: Flash emphasizes throughput and economy where Pro emphasizes peak performance, both under the same low-cost positioning.
DeepSeek-V4-Flash reinforces that high-speed, low-cost tiers can still deliver strong reasoning performance, illustrating that relatively low-cost models can achieve high performance at scale.
Use Cases
Content WritingCode ProgrammingLogical Reasoning