Edit model card

About

weighted/imatrix quants of https://huggingface.co/TroyDoesAI/Codestral-21B-Pruned

static quants are available at https://huggingface.co/mradermacher/Codestral-21B-Pruned-GGUF

Usage

If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi-part files.

Provided Quants

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Link Type Size/GB Notes
GGUF i1-IQ1_S 4.8 for the desperate
GGUF i1-IQ1_M 5.2 mostly desperate
GGUF i1-IQ2_XXS 5.9
GGUF i1-IQ2_XS 6.5
GGUF i1-IQ2_S 6.9
GGUF i1-IQ2_M 7.4
GGUF i1-Q2_K 8.1 IQ3_XXS probably better
GGUF i1-IQ3_XXS 8.4 lower quality
GGUF i1-IQ3_XS 9.0
GGUF i1-Q3_K_S 9.4 IQ3_XS probably better
GGUF i1-IQ3_S 9.5 beats Q3_K*
GGUF i1-IQ3_M 9.8
GGUF i1-Q3_K_M 10.5 IQ3_S probably better
GGUF i1-Q3_K_L 11.4 IQ3_M probably better
GGUF i1-IQ4_XS 11.6
GGUF i1-Q4_0 12.3 fast, low quality
GGUF i1-Q4_K_S 12.3 optimal size/speed/quality
GGUF i1-Q4_K_M 12.9 fast, recommended
GGUF i1-Q5_K_S 14.9
GGUF i1-Q5_K_M 15.3
GGUF i1-Q6_K 17.7 practically like static Q6_K

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):

image.png

And here are Artefact2's thoughts on the matter: https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9

FAQ / Model Request

See https://huggingface.co/mradermacher/model_requests for some answers to questions you might have and/or if you want some other model quantized.

Thanks

I thank my company, nethype GmbH, for letting me use its servers and providing upgrades to my workstation to enable this work in my free time. Additional thanks to @nicoboss for giving me access to his hardware for calculating the imatrix for these quants.

Downloads last month
444
GGUF
Model size
21.5B params
Architecture
llama

1-bit

2-bit

3-bit

4-bit

5-bit

6-bit

Unable to determine this model’s pipeline type. Check the docs .

Quantized from