GGUF models that fit in 4 GB
79024 GGUF language models have at least one quant whose estimated memory use is 4 GB or less at 8K context. Each card shows the largest quant that fits. Sizes run from 1M to 20260731B parameters. The longest trained context is 1280000K tokens.
MI
MiniMax M2.7
Minimax · bartowski
Params230B
Context192K
Q4_K_M138.8 GB
CL
Clef Flash Heretic
Clef · mradermacher
Params8.95B
Context256K
Q4_K_M5.6 GB
QW
GELab-Zero-4B-preview
Qwen · Mungert
Params4B
Context256K
Best fitIQ4_XS 4.0 GB
GR
Groot 19B A3B 261023
Groot · bartowski
Params19B
Context256K
Q4_K_M12.5 GB
OR
Ornith 1.5 35B A3B MTP
Ornith · mradermacher
Params35.5B
Context256K
Q4_K_M21.7 GB
HY
Hy MT2 1.8B
mradermacher
Params1.79B
Context256K
Best fitQ8_0 2.9 GB
MU
multilingual e5 small
Multilingual · cstr
Params118M
Context512
Smallest0.1 GB
TI
Tini Cybersec 8B A1B
Tini · bartowski
Params8.47B
Context125K
Q4_K_M5.2 GB
HO
K2-Horizon-7B
Horizon · bartowski
Params9B
Context512K
Q4_K_M5.8 GB
VE
Veloxy Titan 32B
Veloxy · mradermacher
Params26.9B
Context256K
Q4_K_M16.5 GB
SP
Spark X2.5 4B
Spark · ngquocvinh
Params4.1B
Context1024K
Best fitQ3_K_M 3.9 GB
QW
Qwen2.5 Coder 3B
Qwen · bartowski
Params3.09B
Context32K
Best fitQ6_K_L 3.4 GB
KO
Kolibri 1
Kolibri · bartowski
Params78.1B
Context256K
Q4_K_M49.8 GB
PY
Pyrex 8B Instruct Uncensored i1
Pyrex · mradermacher
Params7.62B
Context32K
Best fitQ2_K 4.0 GB
CO
CohereForAI aya 23 35B
Cohereforai · RichardErkhov
Params35B
Context—
Best fitQ3_K_M 3.4 GB
GR
Granite 3.3 2b Instruct
Granite · ibm-granite
Params2.53B
Context128K
Best fitQ8_0 3.9 GB
PH
phonon 40m pt
Phonon · gradium
Params51M
Context—
Smallest0.1 GB
GU
GUI Owl 1.5 4B Instruct
Gui · mradermacher
Params4B
Context32K
Best fitIQ4_XS 4.0 GB
AS
Asmodeus 24B v3
Asmodeus · mradermacher
Params23.6B
Context128K
Q4_K_M14.3 GB
QW
Qwen3.8 35B A3B Distill
Qwen · mradermacher
Params35.5B
Context256K
Q4_K_M21.7 GB
OR
Ornith-1.5-35B
Ornith · na0x2c6
Params35.5B
Context256K
Q4_K_M21.7 GB
GD
GDP 4B
Gdp · mradermacher
Params4.02B
Context256K
Best fitIQ4_XS 4.0 GB
DE
DeepSeek R1 Distill Qwen 7B
DeepSeek · mradermacher
Params7.62B
Context128K
Best fitQ2_K 4.0 GB
HO
K2 Horizon 32B Heretic
Horizon · mradermacher
Params34.8B
Context512K
Q4_K_M21.1 GB
QW
silicon based girlfriend v2
Qwen · RX5950XT
Params9.2B
Context256K
Q4_K_M5.8 GB
SA
Sage A1.5
Sage · mradermacher
Params35.5B
Context256K
Q4_K_M21.7 GB
CO
Yi Coder 9B Chat
Coder · mradermacher
Params8.83B
Context128K
Best fitIQ2_XS 4.0 GB
QW
Qwen2.5 7B Instruct 1M
Qwen · bartowski
Params7.62B
Context986K
Best fitQ2_K 4.0 GB
GR
Granite-4.1-3B
Granite · unsloth
Params3.4B
Context128K
Best fitQ6_K 4.0 GB
QW
Huihui Qwen3.5 4B Abliterated
Qwen · mradermacher
Params4.21B
Context256K
Best fitIQ4_XS 4.0 GB