DeepSeek V4 Flash (Variant 3)
DeepSeek / 1M
DeepSeek·1M context·Text
Lightweight MoE model with 284B total / 13B active parameters and native 1M context, tuned for low-latency, cost-effective high-concurrency use.
