How much VRAM does Devstral 2 123B Instruct need?

Devstral 2 123B Instruct (123B parameters) requires approximately 99.0 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 Tesla P40 24GB provides.

FWon't run

Too heavy

Using Q4_K_M in vLLM

Capabilities:

Fit status

Too heavy

Decode

3.3 tok/s

TTFT

58308 ms

Safe context

Memory

99.0 GB / 24.0 GB

Weights75.0 GB

KV Cache19.2 GB

Runtime2.4 GB

Headroom2.4 GB

Workload	Grade	Fit	Decode	TTFT	Context
Agentic Coding	F	Too heavy	2.7 tok/s	103500 ms	6K
Chat	F	Too heavy	2.7 tok/s	38812 ms	4K
Coding	F	Too heavy	3.3 tok/s	58308 ms	4K
RAG	F	Too heavy	2.7 tok/s	129375 ms	6K
Reasoning	F	Too heavy	2.7 tok/s	84094 ms	4K

How Devstral 2 123B Instruct (123B params) fits at each quantization level on Tesla P40 24GB (24.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	48.0 GB	Low	F0
Q3_K_S	3	60.3 GB	Low	F0
NVFP4	4

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