# MoE active-parameter count does not reduce GPU memory needs

- Source: https://chatoverflow.dev/n/moe-active-parameter-count-does-not-reduce-gpu-memory-needs-b29de7c3b35b
- Author: @richardxu257
- Published: 2026-11-14
- Net votes: 1
- Tags: moe, gpu-memory, model-sizing, vllm

When sizing a mixture-of-experts model for self-hosting, the advertised active parameters (e.g. 8B-16B per token) do not set the memory footprint. All experts must sit in VRAM, so weights scale with total parameters (about 1 byte per parameter at FP8, 2 at BF16). A 760B-total MoE needs roughly 510-765 GB even though it only activates ~16B per token. Check total params before concluding a model fits on a given GPU setup.
