How much VRAM does Codestral 2 25.08 need?

Codestral 2 25.08 (22B parameters) requires approximately 21.3 GB of memory with Q4_K_M quantization.

What is the best quantization for Codestral 2 25.08?

The recommended quantization for Codestral 2 25.08 is Q4_K_M, which balances quality and memory efficiency.

Can it run?

Yes, RX 7900 XT 20GB can run Codestral 2 25.08 with a C grade (Runs with offload (needs ~0.8 GB host RAM)). Expected decode speed: 34.0 tok/s.

CUsable

Runs with offload (needs ~0.8 GB host RAM)

Using Q4_K_M in vLLM

Capabilities:

Fit status

Runs with offload (needs ~0.8 GB host RAM)

Decode

34.0 tok/s

TTFT

5699 ms

Safe context

15K

Memory

21.3 GB / 20.0 GB

Offload

10%

Weights13.4 GB

KV Cache3.4 GB

Runtime2.4 GB

Headroom2.0 GB

Workload	Grade	Fit	Decode	TTFT	Context
Agentic Coding	F	Too heavy	35.8 tok/s	7873 ms	26K
Chat	C	Runs with offload	35.8 tok/s	2952 ms	8K
Coding	C	Runs with offload (needs ~0.8 GB host RAM)	34.0 tok/s	5699 ms	15K
RAG	F	Too heavy	35.8 tok/s	9841 ms	26K
Reasoning	C	Runs with offload (needs ~0.8 GB host RAM)	34.0 tok/s	6736 ms

How Codestral 2 25.08 (22B params) fits at each quantization level on RX 7900 XT 20GB (20.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	8.6 GB	Low	D39
Q3_K_S	3	10.8 GB	Low	C41
NVFP4	4

HuggingFace

huggingface-cli download codestral-2-25.08

Upgrade options

MacBook Pro M4 32GBBest value

C6.4 tok/s decode

~$799 MSRP

RX 7900 XTX 24GBBudget pick

C51.5 tok/s decode

~$999 MSRP

RTX 5090 32GBBiggest leap