How much VRAM does Devstral 2 123B Instruct need?

Devstral 2 123B Instruct (123B parameters) requires approximately 104.6 GB of memory with Q4_K_M quantization.

What is the best quantization for Devstral 2 123B Instruct?

The recommended quantization for Devstral 2 123B Instruct is Q4_K_M, which balances quality and memory efficiency.

Can it run?

No, Devstral 2 123B Instruct requires more memory than NVIDIA H100 80GB provides.

FWon't run

Too heavy

Using Q4_K_M in vLLM

Capabilities:

Fit status

Too heavy

Decode

32.1 tok/s

TTFT

6022 ms

Safe context

12K

Memory

104.6 GB / 80.0 GB

Weights75.0 GB

KV Cache19.2 GB

Runtime2.4 GB

Headroom8.0 GB

Workload	Grade	Fit	Decode	TTFT	Context
Agentic Coding	F	Too heavy	32.1 tok/s	8760 ms	21K
Chat	D	Very compromised	27.8 tok/s	3795 ms	7K
Coding	F	Too heavy	32.1 tok/s	6022 ms	12K
RAG	F	Too heavy	32.1 tok/s	10950 ms	21K
Reasoning	F	Too heavy	32.1 tok/s	7117 ms	12K

How Devstral 2 123B Instruct (123B params) fits at each quantization level on NVIDIA H100 80GB (80.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	48.0 GB	Low	C42
Q3_K_SBest for your GPU	3	60.3 GB	Low	C45

Upgrade options

MacBook Pro M3 Max 128GBBest value

D3 tok/s decode

~$2,499 MSRP

AMD Instinct MI350X 288GBBudget pick

C77.8 tok/s decode

~$8,000 MSRP

NVIDIA B200 180GBBiggest leap