GGUF models that fit in 48 GB

201364 GGUF language models have at least one quant whose estimated memory use is 48 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.

Most downloadedNewestLargestLongest contextSmallest
QW
Qwen2.5 7B Instruct Abliterated v2
Qwen · lumen-models
Params7.62B
Context32K
Best fitQ4_K_M 5.7 GB
GGUF
QW
Qwen3.6 35B A3B MTP APEX I MiniPlus V2.1
Qwen · IsValorum
Params35.5B
Context256K
Smallest1.2 GB
ReasoningTool useVision
HU
HunyuanOCR
Hunyuanocr · ggml-org
Params539M
Context32K
Best fitBF16 2.4 GB
Vision
MO
moondream2
Moondream · ggml-org
Params1.42B
Context2K
Best fitF16 5.0 GB
Vision
QW
Qwen3 4B Instruct 2507 Polaris Alpha Distill Heretic Abliterated
Qwen · mradermacher
Params4B
Context256K
Best fitF16 9.8 GB
GGUF
IM
Z-Image-Engineer-V6-SMART-DoRA-StyleBlendV9-25V17-75
Image · BennyDaBall
Params4B
Context40K
Best fitF16 9.8 GB
Coding
QW
Qwen3.8 27B BaseQuant V1
Qwen · basecompute
Params27.3B
Context256K
Smallest8.4 GB
ReasoningTool use
RW
Rwkv 6 World 7b
Rwkv · latestissue
Params7.64B
Context1024K
Best fitF32 34.7 GB
GGUF
KE
Kev 9B
Kev · ggml-org
Params9B
Context256K
Best fitBF16 19.5 GB
GGUF
QW
dolphin-2.9.2-qwen2-72b
Qwen · bartowski
Params72.7B
Context128K
Best fitIQ4_XS 42.9 GB
CodingTool use
HO
K2-Horizon-32B
Horizon · bartowski
Params34.8B
Context512K
Best fitQ8_0 39.6 GB
GGUF
QW
Huihui Qwen3.5 0.8B abliterated claude mythos distilled 25k
Qwen · azminawwar
Params773M
Context256K
Best fitBF16 2.5 GB
ReasoningTool useVision
QW
Qwen3.5 27B Ultra Uncensored Heretic v2
Qwen · mradermacher
Params27B
Context256K
Best fitQ6_K 24.7 GB
GGUF
GO
Goetia 26B A4B v1.3 Absolute Heretic ARA
Goetia · mradermacher
Params25.2B
Context256K
Best fitQ6_K 31.2 GB
Roleplay
QW
Qwen3.8 Flash Next Uncensored UD Q4 K XL
Qwen · NeosQ
Params177B
Context256K
Best fitQ4_K_XL 5.5 GB
ReasoningTool useVision
DR
Drex v1.5
Drex · nace-ai
Params9B
Context256K
Best fitQ8_0 11.1 GB
GGUF
LL
tap-m_luna-ai-llama2-uncensored
Llama · TheBloke
Params6.74B
Context2K
Best fitQ8_0 12.0 GB
GGUF
GR
Granite-4.0-H-Tiny
Granite · unsloth
Params6.94B
Context1024K
Best fitBF16 16.4 GB
GGUF
QW
Qwen3 VL 8B Instruct Uncensored V2
Qwen · mradermacher
Params8.19B
Context256K
Best fitF16 18.1 GB
Vision
QW
Qwen2.5 14B Instruct 1M
Qwen · lmstudio-community
Params14.8B
Context986K
Best fitQ8_0 17.8 GB
GGUF
MI
dolphin 2.8 mistral 7b v02
Mistral · lmstudio-community
Params7.24B
Context32K
Best fitQ8_0 9.3 GB
CodingTool use
PY
Pyrut 8B Instruct Uncensored i1
Pyrex · mradermacher
Params7.62B
Context32K
Best fitQ6_K 7.2 GB
CodingTool use
RO
RoLlama2-7b-Instruct
Rollama · legraphista
Params6.74B
Context4K
Best fitBF16 18.3 GB
GGUF
AP
Apex 35B V1.0
Apex · bartowski
Params35.5B
Context256K
Best fitQ8_0 39.0 GB
ReasoningTool useVision
SW
Swish 31B Final
Swish · mradermacher
Params30.7B
Context256K
Best fitQ3_K_M 48.0 GB
ReasoningTool use
QW
Qwen2.5 Coder 0.5B Instruct
Qwen · bartowski
Params494M
Context32K
Best fitF16 1.6 GB
Coding
GE
Gemma 4 31B Monadize v0.1
Gemma · mradermacher
Params30.7B
Context256K
Best fitQ3_K_M 48.0 GB
ReasoningTool use
PH
phonon 40m es
Phonon · gradium
Params51M
Context—
Smallest0.1 GB
GGUF
GE
gemma 3 4b it qat q4 0
Gemma · google
Params3.88B
Context—
Best fitQ4_0 4.0 GB
Vision
IN
Ingot VI 27B Q2
Ingot · dendriteholdings
Params26.9B
Context256K
Smallest7.2 GB
ReasoningTool use

Other memory sizes

Keep browsing