Text model · DeepSeek-V2

Can I run DeepSeek-V2-Lite? Pick your device.

DeepSeek-V2-Lite is a 16B MoE (2.4B active) model. At Q4_K_M the weights are ~10.4 GB; with KV cache and runtime overhead, budget ~12.2 GB. Select your hardware below to see the full compatibility breakdown.

Memory figures are estimates. See methodology.