How much VRAM does Devstral Small 1.1 need?

Devstral Small 1.1 (24B parameters) requires approximately 21.6 GB of memory with Q4_K_M quantization.

What is the best quantization for Devstral Small 1.1?

The recommended quantization for Devstral Small 1.1 is Q4_K_M, which balances quality and memory efficiency.

Can it run?

Yes, RTX 4000 Ada 20GB can run Devstral Small 1.1 with a D grade (Runs with offload). Expected decode speed: 16.6 tok/s.

DPoor

Runs with offload

Using Q4_K_M in Ollama

Capabilities:

Fit status

Runs with offload

Decode

16.6 tok/s

TTFT

11665 ms

Safe context

15K

Memory

21.6 GB / 20.0 GB

Offload

10%

Weights14.6 GB

KV Cache3.8 GB

Runtime1.2 GB

Headroom2.0 GB

Workload	Grade	Fit	Decode	TTFT	Context
Agentic Coding	F	Too heavy	19.2 tok/s	14682 ms	25K
Chat	C	Runs with offload	19.2 tok/s	5506 ms	8K
Coding	D	Runs with offload	16.6 tok/s	11665 ms	15K
RAG	F	Too heavy	19.2 tok/s	18352 ms	25K
Reasoning	D	Runs with offload (needs ~1.1 GB host RAM)	18.0 tok/s	12725 ms	15K

How Devstral Small 1.1 (24B params) fits at each quantization level on RTX 4000 Ada 20GB (20.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	9.4 GB	Low	D39
Q3_K_S	3	11.8 GB	Low	C42
NVFP4	4

Ollama

ollama run devstral-small-2507

HuggingFace

huggingface-cli download devstral-small-2507

Upgrade options

MacBook Pro M4 32GBBest value

C5.9 tok/s decode

~$799 MSRP

RX 7900 XTX 24GBBudget pick

C47.2 tok/s decode

~$999 MSRP

RTX 5090 32GBBiggest leap