mradermacher/DeepSeek-R1-i1-GGUF

About

weighted/imatrix quants of https://huggingface.co/deepseek-ai/DeepSeek-R1

static quants are available at https://huggingface.co/mradermacher/DeepSeek-R1-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
PART 1 PART 2 PART 3	i1-IQ1_S	133.7	for the desperate
PART 1 PART 2 PART 3 PART 4	i1-IQ1_M	149.0	mostly desperate
PART 1 PART 2 PART 3 PART 4	i1-IQ2_XXS	174.5
PART 1 PART 2 PART 3 PART 4	i1-IQ2_XS	195.2
PART 1 PART 2 PART 3 PART 4	i1-IQ2_S	197.1
P1 P2 P3 P4 P5	i1-IQ2_M	217.5
P1 P2 P3 P4 P5	i1-Q2_K_S	224.8	very low quality
P1 P2 P3 P4 P5	i1-Q2_K	244.1	IQ3_XXS probably better
P1 P2 P3 P4 P5 P6	i1-IQ3_XXS	258.0	lower quality
P1 P2 P3 P4 P5 P6	i1-IQ3_XS	272.9
P1 P2 P3 P4 P5 P6	i1-IQ3_S	289.2	beats Q3_K*
P1 P2 P3 P4 P5 P6	i1-Q3_K_S	289.2	IQ3_XS probably better
P1 P2 P3 P4 P5 P6	i1-IQ3_M	292.2
P1 P2 P3 P4 P5 P6 P7	i1-Q3_K_M	319.3	IQ3_S probably better
P1 P2 P3 P4 P5 P6 P7 P8	i1-Q3_K_L	347.5	IQ3_M probably better
P1 P2 P3 P4 P5 P6 P7 P8	i1-IQ4_XS	357.2
P1 P2 P3 P4 P5 P6 P7 P8	i1-Q4_0	379.1	fast, low quality
P1 P2 P3 P4 P5 P6 P7 P8	i1-Q4_K_S	380.1	optimal size/speed/quality
P1 P2 P3 P4 P5 P6 P7 P8 P9	i1-Q4_K_M	404.5	fast, recommended
P1 P2 P3 P4 P5 P6 P7 P8 P9	i1-Q4_1	420.0
P1 P2 P3 P4 P5 P6 P7 P8 P9 P10	i1-Q5_K_S	461.9
P1 P2 P3 P4 P5 P6 P7 P8 P9 P10	i1-Q5_K_M	475.5
P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12	i1-Q6_K	550.9	practically like static Q6_K

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

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 private supercomputer, enabling me to provide many more imatrix quants, at much higher quality, than I would otherwise be able to.