GGUF models that fit in 12 GB
171517 GGUF language models have at least one quant whose estimated memory use is 12 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 4194304K tokens.
QW
Bonsai 2 27B 1bit CRACK
Qwen · dealignai
Params27B
Context256K
Smallest5.9 GB
QW
Qwen3.5-9B
Qwen · jc-builds
Params8.95B
Context256K
Best fitQ4_K_M 7.3 GB
QW
Qwen2.5 Coder 32B Instruct
Qwen · CISCai
Params32.8B
Context32K
Best fitIQ2_XXS 11.7 GB
DA
L3 DARKEST PLANET 16.5B
Darkest · DavidAU
Params16.5B
Context8K
Best fitQ3_K_M 11.2 GB
QW
Qwen3.8 27B Optimal Ablit Heretic
Qwen · mradermacher
Params27B
Context256K
Best fitIQ2_S 12.0 GB
EU
EuroLLM 1.7B Instruct
Eurollm · mradermacher
Params1.66B
Context8K
Best fitF16 4.6 GB
QW
Qwen2.5 VL 32B Instruct
Qwen · bartowski
Params32.8B
Context125K
Best fitIQ2_XXS 11.7 GB
QW
Qwen3.6 14B A3B VibeForged v2
Qwen · tvall43
Params13.8B
Context256K
Best fitMXFP4_MOE 9.8 GB
QW
Qwen3 30B A3B Thinking 2507
Qwen · bartowski
Params30.5B
Context256K
Best fitIQ2_M 11.2 GB
QW
Qwen3 4b Heretic
Qwen · DreamFast
Params11.8B
Context40K
Best fitQ6_K 11.4 GB
FA
Fara1.5 27B
Fara · bartowski
Params27B
Context256K
Q4_K_M17.5 GB
GU
GUI Owl 1.5 8B Instruct
Gui · mradermacher
Params8.19B
Context32K
Best fitQ8_0 10.4 GB
QW
Huihui Qwen3.5 27B Abliterated
Qwen · mradermacher
Params27B
Context256K
Best fitIQ2_M 12.0 GB
VI
Vikhr Nemo 12B Instruct R 21 09 24 Q4 K M
Vikhr · VlSav
Params12.2B
Context1000K
Best fitQ4_K_M 9.3 GB
SW
Swish 26B A4B Final
Swish · mradermacher
Params25.2B
Context256K
Q4_K_M16.8 GB
GP
Gpt Oss 20b Heretic
gpt-oss · bartowski
Params20.9B
Context128K
Q4_K_M15.9 GB
QW
Qwen3.8 27B Antislop
Qwen · mradermacher
Params27B
Context256K
Best fitIQ2_S 12.0 GB
QW
Qwen3.8-Flash-Next
Qwen · ji-farthing
Params177B
Context256K
Smallest2.4 GB
PY
Pythia 6.9b
Pythia · mradermacher
Params6.86B
Context2K
Best fitQ6_K 10.4 GB
GP
Gpt Oss 120b Abliterated
gpt-oss · mradermacher
Params117B
Context128K
Q4_K_M87.9 GB
SU
SuperQwen3.8-27b-abliterated
Superqwen · Jiunsong
Params27B
Context256K
Best fitQ4_0 4.3 GB
DE
deepseek-ai_deepseek-coder-1.3b-base
DeepSeek · TheBloke
Params1.35B
Context16K
Best fitQ8_0 3.5 GB
LL
Dolphin 3.0 Llama 3.2 1b
Llama · bartowski
Params1.24B
Context128K
Best fitF32 5.7 GB
LF
LFM2.5 24B A2B
Lfm · mradermacher
Params24B
Context128K
Best fitQ2_K 12.0 GB
QW
Qwen3.5 122B A10B Abliterated v1
Qwen · mradermacher
Params122B
Context256K
Q4_K_M74.2 GB
QW
Qwen2.5 Coder 32B Instruct 3MPER0RR Abliterated
Qwen · mradermacher
Params32.8B
Context32K
Best fitIQ2_XXS 11.7 GB
MA
UI Mate 27B
Mate · bartowski
Params27B
Context256K
Best fitIQ2_XXS 11.8 GB
HE
HERO 9B
Hero · mradermacher
Params8.95B
Context256K
Best fitQ6_K 8.9 GB
SM
SmolVLM2-256M-Video-Instruct
Smolvlm · ggml-org
Params163M
Context8K
Best fitBF16 1.0 GB
QW
Qwen2.5 1.5b
Qwen · ElementMerc
Params1.54B
Context32K
Best fitF16 3.8 GB