Model

DeepSeek-V4-Flash

DeepSeek's cost-efficient open-weight MoE model (284B total / 13B active) for high-throughput reasoning and coding, with a 1M-token context under an MIT license.

Open sourcereasoningDeepSeek-V4open weightReleased 2026-04-241M contextDeepSeekOfficial page

Model metadata

Modalities: text

Recent coverage

Related events

Related papers