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.
QW
Qwen2-1.5B
Qwen · legraphista
Params1.54B
Context128K
Best fitBF16 3.8 GB
BO
prism ml Bonsai 8B unpacked
Bonsai · bartowski
Params8.19B
Context64K
Q4_K_M5.2 GB
LL
Llama 3.2 1b Abliterated
Llama · mradermacher
Params1.24B
Context128K
Best fitQ6_K 1.8 GB
LL
Reflection Llama 3.1 8B i1
Llama · mradermacher
Params8B
Context128K
Best fitIQ2_XXS 4.0 GB
LY
LYNX ALPHA
Lynx · mradermacher
Params8.95B
Context256K
Q4_K_M5.6 GB
LF
Lfm2.5 2.6b Fable5 Coding Agent Heretic
Lfm · mradermacher
Params2.7B
Context128K
Best fitQ3_K_L 4.0 GB
VN
Vnpen Writer 9b v1.0
Vnpen · mradermacher
Params8.95B
Context256K
Q4_K_M5.6 GB
DE
DeepSeek R1 Distill Qwen 7B Abliterated v2
DeepSeek · mradermacher
Params7.62B
Context128K
Best fitQ2_K 4.0 GB
LL
Llama Guard 3 8B
Llama · legraphista
Params8B
Context128K
Best fitIQ2_XXS 4.0 GB
DA
DataPilot ArrowPro 7B KUJIRA
Datapilot · mmnga
Params7.24B
Context4K
Best fitIQ2_S 3.9 GB
ST
L3-8B-Stheno-v3.3-32K
Stheno · DavidAU
Params8B
Context32K
Best fitIQ1_M 3.7 GB
CO
Cosmos Reason2 2b
Cosmos · apolo13x
Params2.03B
Context256K
Best fitQ8_0 3.6 GB
ST
Starling-LM-7B-alpha-x2
Starling · mradermacher
Params7.24B
Context8K
Best fitIQ2_S 3.9 GB
GR
Granite 4.0 H Tiny
Granite · bartowski
Params6.94B
Context1024K
Q4_K_M4.3 GB
PY
Pythia 1b
Pythia · mradermacher
Params1.01B
Context2K
Best fitQ6_K 2.4 GB
PR
Privacy Filter
Privacy · LocalAI-io
Params1.4B
Context128K
Best fitF16 3.4 GB
GE
Gemma-4-E4B-It
Gemma · ente-ai
Params7.5B
Context128K
Best fitBF16 2.1 GB
AS
AstaBrief_8B
Astabrief · mradermacher
Params8.2B
Context40K
Best fitIQ1_M 4.0 GB
MI
MiniCPM5 2B DSpark
MiniCPM · openbmb
Params324M
Context128K
Smallest0.7 GB
QW
Qwen3.6 35B A3B Uncensored Heretic Native MTP Preserved
Qwen · mradermacher
Params35.5B
Context256K
Q4_K_M21.7 GB
SP
Spark X2.6 4B Uncensored GSQ RCO
Spark · abenzerps
Params4.11B
Context1024K
Best fitIQ3_M 4.0 GB
QW
Qwen3.8 4B Heretic
Qwen · mradermacher
Params4.21B
Context256K
Best fitQ3_K_L 4.0 GB
PY
Pythia 410m Deduped
Pythia · mradermacher
Params405M
Context2K
Best fitQ4_0 1.6 GB
QW
Qwen3.8 2B Distill Heretic
Qwen · mradermacher
Params1.88B
Context256K
Best fitQ6_K 2.5 GB
MI
Voxtral Small 24B 2507
Mistral · bartowski
Params23.6B
Context128K
Q4_K_M14.3 GB
AU
Audar ASR V1 Turbo
Audar · audarai
Params2.03B
Context64K
Best fitQ8_0 3.6 GB
QW
Qwen3 8B Socratic SFT
Qwen · mradermacher
Params8.19B
Context40K
Best fitIQ1_M 4.0 GB
GE
Skaldic Gemma 4 E4B
Gemma · mradermacher
Params7.46B
Context128K
Q4_K_M5.3 GB
QW
Qwen2-1.5B-Instruct
Qwen · legraphista
Params1.54B
Context32K
Best fitBF16 3.8 GB
JA
Jais-3-7B
Jais · mradermacher
Params9B
Context512K
Q4_K_M5.6 GB