Question 1

Can Gaudi 3 128GB run Llama 4 Scout 17B 16E?

Accepted Answer

Yes, Gaudi 3 128GB can run Llama 4 Scout 17B 16E with a B grade (Runs well). Expected decode speed: 94.3 tok/s.

Question 2

How much VRAM does Llama 4 Scout 17B 16E need?

Accepted Answer

Llama 4 Scout 17B 16E (109B parameters) requires approximately 83.1 GB of memory with Q4_K_M quantization.

Question 3

What is the best quantization for Llama 4 Scout 17B 16E?

Accepted Answer

The recommended quantization for Llama 4 Scout 17B 16E is Q4_K_M, which balances quality and memory efficiency.

Workload	Grade	Fit	Decode	TTFT	Context
Agentic Coding	B	Runs well	91.7 tok/s	3071 ms	48K
Chat	B	Runs well	91.7 tok/s	1152 ms	13K
Coding	B	Runs well	94.3 tok/s	2052 ms	25K
RAG	B	Runs well	91.7 tok/s	3838 ms	48K
Reasoning	B	Runs well	91.7 tok/s	2495 ms	25K

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	42.5 GB	Low	D36
Q3_K_S	3	53.4 GB	Low	D38
NVFP4	4

Can Gaudi 3 128GB run Llama 4 Scout 17B 16E?

Memory breakdown