Llama400B params17B active (MoE)10000k contextpopular

Llama 4 Maverick 17Bx128 locally

Llama 4 Maverick 17Bx128 is an open-weight LLM from the Llama family with 400B parameters. Main use: chat, RAG and general assistance. Detected minimum hardware: Instinct MI325X (256 GB).

Technical facts
Parameters400B
Max context10000k
Q4_K_M251.5 GB
Q5_K_M307.3 GB
Q8447.0 GB
FP16894.1 GB
FamilyLlama
Last sync2026-05-12

Available quantizations

Q4_K_M
251.5GB

Acceptable. Good compromise when VRAM is limited.

Q5_K_M
307.3GB

Good quality. Sweet spot for size and precision.

Q8
447.0GB

Near-FP16 quality. Comfortable for production.

FP16
894.1GB

Reference precision. Maximum quality, doubled VRAM.

Compatible GPUs

GPUs that can run Llama 4 Maverick 17Bx128 on a single card, ranked by VRAM headroom.

Similar models

VRAM estimates: parameters x bits/8 plus margin. Real performance varies by engine, context length and batch size.
sync: 2026-05-12