How much VRAM does Codestral-22B-v0.1-hf need?

Q4_K_M is exactly 13,341,237,632 bytes (12.42 GiB) in weights. Add the KV cache, which depends on context length, plus roughly half a gigabyte of runtime overhead.

How large is Codestral-22B-v0.1-hf's KV cache?

7.00 GiB at 32K context with an f16 cache, computed per layer from the model's architecture. Quantizing the cache to q8_0 roughly halves it.

bullerwins · text

Codestral-22B-v0.1-hf

bullerwins/Codestral-22B-v0.1-hf

Codestral-22B-v0.1-hf at Q4_K_M is exactly 13,341,237,632 bytes (12.42 GiB / 13.34 GB) — an effective 4.797 bits per weight, not the nominal 4. Its KV cache at 32K is 7.00 GiB.

From the file· summed from 1 file(s)From the file· KV per layer

Parameters

22.2B

Architecture

llama

56 layers

Context

32,768

native (config.json)

License

—

Shipped quantizations

● exact bytes, summed from published files

Quant	Size●	Exact bytes●	Effective bpw●	Tensors●	Publisher
IQ1_S	4.50 GiB	4,829,487,488	1.737	—	MaziyarPanahi
IQ1_M	4.91 GiB	5,267,136,896	1.894	—	MaziyarPanahi
IQ2_XS	6.19 GiB	6,646,145,408	2.390	—	MaziyarPanahi
Q2_K	7.70 GiB	8,272,093,568	2.975	—	MaziyarPanahi
IQ3_XS	8.55 GiB	9,176,097,152	3.300	—	MaziyarPanahi
Q3_K_S	8.98 GiB	9,641,271,680	3.467	—	MaziyarPanahi
Q3_K_M	10.02 GiB	10,756,825,472	3.868	—	MaziyarPanahi
Q3_K_L	10.92 GiB	11,730,428,288	4.218	—	MaziyarPanahi
IQ4_XS	11.12 GiB	11,935,293,824	4.292	—	MaziyarPanahi
Q4_K_S	11.79 GiB	12,660,384,128	4.553	—	MaziyarPanahi
Q4_K_M	12.42 GiB	13,341,237,632	4.797	—	MaziyarPanahi
Q5_K_S	14.27 GiB	15,324,815,744	5.511	—	MaziyarPanahi
Q5_K_M	14.64 GiB	15,722,553,728	5.654	—	MaziyarPanahi
Q6_K	17.00 GiB	18,252,702,080	6.564	—	MaziyarPanahi
Q8_0	22.02 GiB	23,640,547,712	8.501	—	MaziyarPanahi

KV cache by context

computed per layer

Context	KV cache (f16)●	Flat formula	Overstated by	Full / windowed / recurrent
4,096	0.88 GiB	0.88 GiB	—	56 / 0 / 0
8,192	1.75 GiB	1.75 GiB	—	56 / 0 / 0
16,384	3.50 GiB	3.50 GiB	—	56 / 0 / 0
32,768	7.00 GiB	7.00 GiB	—	56 / 0 / 0
65,536	14.00 GiB	14.00 GiB	—	56 / 0 / 0
131,072	28.00 GiB	28.00 GiB	—	56 / 0 / 0

Compare with

same modality, comparable size

vs Qwen3-Coder-30B-A3B-Instruct vs Qwen3.6-27B vs Qwen3.8-27B vs gemma-4-12B-it-qat-q4_0-unquantized vs gemma-4-E4B-it vs Qwen3-30B-A3B-Thinking-2507 All alternatives →Will it fit on my card? →

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 11.65 GiB. The real file is 12.42 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

from config.json

Layers

Attention heads

KV heads

Head dim

128

Hidden size

6144

Vocab

32,768

Sliding window

none

SWA period

—

MLA

Experts

—

Experts per token

—

use_sliding_window

—

Questions people ask

How much VRAM does Codestral-22B-v0.1-hf need?: Q4_K_M is exactly 13,341,237,632 bytes (12.42 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
How large is Codestral-22B-v0.1-hf's KV cache?: 7.00 GiB at 32K context with an f16 cache, computed per layer. Quantizing the cache to q8_0 roughly halves it, which is often the difference between a context length fitting and not.
Which quantization of Codestral-22B-v0.1-hf should I use?: Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.