How much VRAM does Codestral 2 25.08 need?

Codestral 2 25.08 (22B parameters) requires approximately 20.5 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?

No, Codestral 2 25.08 requires more memory than RTX 4070 12GB provides.

FWon't run

Too heavy

Using Q4_K_M in vLLM

Capabilities:

Fit status

Too heavy

Decode

27.0 tok/s

TTFT

7160 ms

Safe context

Memory

20.5 GB / 12.0 GB

Weights13.4 GB

KV Cache3.4 GB

Runtime2.4 GB

Headroom1.2 GB

Workload	Grade	Fit	Decode	TTFT	Context
Agentic Coding	F	Too heavy	27.0 tok/s	10414 ms	16K
Chat	F	Too heavy	27.0 tok/s	3905 ms	5K
Coding	F	Too heavy	27.0 tok/s	7160 ms	9K
RAG	F	Too heavy	27.0 tok/s	13018 ms	16K
Reasoning	F	Too heavy	27.0 tok/s	8461 ms	9K

How Codestral 2 25.08 (22B params) fits at each quantization level on RTX 4070 12GB (12.0 GB usable).

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	8.6 GB	Low	C44
Q3_K_S	3	10.8 GB	Low	C44
NVFP4	4

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