Download ranking
Most-downloaded local LLMs that run in 32 GB of VRAM
Only models with a quantization whose estimated memory fits in 32 GB at an 8,192-token context, in the same order. The quantization shown is the largest (most precise) one that fits.
Top 50 of 724 models
Rank changes are compared with the record from Nov 4, 2026.
-
1 no changeQwen3.8-27B
Qwen 27.3B
Downloads (30 days) 10,491,667
Likes 6,515
UD-Q8_K_XL 33.2 GB
-
2 no changeOrnith-1.5-9B
ornith-ai 9.2B
Downloads (30 days) 5,801,033
Likes 743
BF16 19.8 GB
-
3 4up 4Huihui-Qwen3.6-35B-A3B-abliterated-MTP
huihui-ai 35.5B
Downloads (30 days) 4,969,070
Likes 88
Q6_K 30.5 GB
-
4 1down 1Ornith-1.5-35B-A3B
ornith-ai 35.5B
Downloads (30 days) 4,267,948
Likes 721
Q6_K_L 32.1 GB
-
5 1down 1Huihui-Qwen3.8-27B-abliterated
huihui-ai 27.3B
Downloads (30 days) 3,958,526
Likes 1,907
UD-Q8_K_XL 33.2 GB
-
6 newOTel-2.0-Decision-26B-A4B-4bit
farbodtavakkoli 25.2B
Downloads (30 days) 3,375,370
Likes 7
Q8_0 28.4 GB
-
7 2down 2Qwen3.8-27B-iMatrix-NVFP4-MTP
cdiamond
Downloads (30 days) 3,253,089
Likes 48
NVFP4 18.9 GB
-
8 2down 2Tiel-Coder-35B-A3B-GGUF-MTP
peculiar-ragdoll
Downloads (30 days) 3,055,106
Likes 355
UD-Q6_K_XL 33.5 GB
-
9 1up 1gemma-4-12B-it
google 11.9B
Downloads (30 days) 2,834,643
Likes 1,158
BF16 25.5 GB
-
10 1down 1Qwen3.5-9B
Qwen 9B
Downloads (30 days) 2,657,097
Likes 1,270
BF16 19.8 GB
-
11 3down 3Qwen3.8-27B-Uncensored
JonathanColetti 27.3B
Downloads (30 days) 2,472,728
Likes 1,964
Q8_0 30.8 GB
-
12 1up 1MOSS-Music-8B-Instruct
OpenMOSS-Team
Downloads (30 days) 2,446,681
Likes 11
F16 18.4 GB
-
13 1up 1gemma-4-26B-A4B-it
google 25.2B
Downloads (30 days) 2,434,395
Likes 1,579
UD-Q8_K_XL 29.2 GB
-
14 3down 3Ornith-1.0-9B
ornith-ai 9B
Downloads (30 days) 2,374,819
Likes 699
BF16 19.3 GB
-
15 1up 1Qwen3.6-35B-A3B
Qwen 34.7B
Downloads (30 days) 2,265,195
Likes 2,020
UD-Q6_K_XL 33.1 GB
-
16 1up 1Qwen3-VL-8B-Instruct
Qwen 8.2B
Downloads (30 days) 2,195,718
Likes 323
BF16 18.4 GB
-
17 4up 4Qwen3.6-35B-A3B-MTP
unsloth 35.5B
Downloads (30 days) 2,170,348
Likes 1,046
UD-Q6_K_XL 33.9 GB
-
18 3down 3
Downloads (30 days) 2,097,725
Likes 1,514
BF16 19.8 GB
-
19 7down 7
Downloads (30 days) 1,997,118
Likes 2,363
Q8_K 33.2 GB
-
20 1down 1Qwen3.5-4B
Qwen 4.2B
Downloads (30 days) 1,980,253
Likes 673
BF16 10.0 GB
-
21 1down 1Ornith-1.5-9B-uncensored
junafinity 9B
Downloads (30 days) 1,736,789
Likes 278
F16 19.3 GB
-
22 4down 4Ternary-Bonsai-2-27B
prism-ml 26.9B
Downloads (30 days) 1,620,572
Likes 2,982
Q2_0 9.37 GB
-
23 1down 1Swift-1.5-Qwen3.8-27B-GSQ-RCO
ukisai 26.9B
Downloads (30 days) 1,607,683
Likes 508
IQ3_S 13.9 GB
-
24 2up 2gemma-4-E4B-it
google 7.5B
Downloads (30 days) 1,525,316
Likes 948
BF16 16.3 GB
-
25 2down 2Ornith-1.0-35B
ornith-ai 34.7B
Downloads (30 days) 1,488,276
Likes 1,064
Q6_K 29.8 GB
-
26 2down 2
Downloads (30 days) 1,459,029
Likes 2,828
Q8_0 32.0 GB
-
27 2down 2Dirk-Qwen3.8-27B
peculiar-ragdoll 27.3B
Downloads (30 days) 1,400,070
Likes 265
UD-Q8_K_XL 33.2 GB
-
28 1up 1Wan2.1
DeepBeepMeep
Downloads (30 days) 1,397,059
Likes 73
IQ4_NL 30.6 GB
-
29 2down 2Qwen3.8-27B-GSQ-RCO
ISTA-DASLab 26.9B
Downloads (30 days) 1,364,705
Likes 2,618
IQ3_S 13.9 GB
-
30 no changeQwen3.8-35B-A3B-Distill
empero-ai 35.5B
Downloads (30 days) 1,336,101
Likes 276
Q6_K 30.5 GB
-
31 2up 2Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive
HauhauCS 34.7B
Downloads (30 days) 1,332,334
Likes 4,137
Q6_K 31.9 GB
-
32 4down 4Qwen3-4B
Qwen 4B
Downloads (30 days) 1,320,091
Likes 505
BF16 10.1 GB
-
33 2down 2gemma-4-E4B-it-qat-q4_0
google 7.5B
Downloads (30 days) 1,297,573
Likes 186
Q4_0 6.40 GB
-
34 24up 24Qwen3.8-Flash-Next-Gyro
agentionai 125.7B
Downloads (30 days) 1,243,631
Likes 181
TQ1_0 31.1 GB
-
35 22up 22Ternary-Bonsai-2-9B
prism-ml
Downloads (30 days) 1,219,968
Likes 163
Q8_0 4.04 GB
-
36 4down 4HyperCLOVAX-SEED-Text-Instruct-1.5B-Q4_K_M
rippertnt 1.8B
Downloads (30 days) 1,164,155
Likes 17
Q4_K_M 2.70 GB
-
37 1up 1gemma-4-12B-it-qat-q4_0
google 11.9B
Downloads (30 days) 1,164,140
Likes 396
Q4_0 8.69 GB
-
38 2down 2Hy-MT2-7B
tencent 7.5B
Downloads (30 days) 1,149,223
Likes 103
Q8_0 9.86 GB
-
39 25up 25Qwen3.8-Flash-Next-Gyro-abliterated-Huihui
agentionai 125.7B
Downloads (30 days) 1,147,499
Likes 54
TQ1_0 31.1 GB
-
40 9up 9Swift-1.5-Qwen3.8-27B-heretic-GSQ-RCO
jegreene 27.3B
Downloads (30 days) 1,142,899
Likes 33
IQ3_S 13.9 GB
-
41 7down 7Qwen3.8-27B-Heretic-Abliterated-Uncensored
0bserverx 26.9B
Downloads (30 days) 1,135,253
Likes 869
Q8_0 30.8 GB
-
42 7down 7Qwen3-VL-4B-Instruct
Qwen 4B
Downloads (30 days) 1,131,643
Likes 171
BF16 10.1 GB
-
43 6down 6Muse-Glimmer-30B
meta-models 27.9B
Downloads (30 days) 1,123,221
Likes 981
UD-Q8_K_XL 33.5 GB
-
44 4down 4gemma-4-E2B-it
google 4.6B
Downloads (30 days) 1,114,000
Likes 513
BF16 10.4 GB
-
45 6down 6gpt-oss-20b
openai 20.9B
Downloads (30 days) 1,103,373
Likes 1,192
F16 15.0 GB
-
46 5down 5Qwen3-8B
Qwen 8.2B
Downloads (30 days) 1,085,620
Likes 594
BF16 18.4 GB
-
47 22up 22Ternary-Bonsai-2-2B
prism-ml 1.9B
Downloads (30 days) 1,049,409
Likes 53
Q2_0 1.71 GB
-
48 4up 4
Downloads (30 days) 1,036,784
Likes 294
Q8_0 32.0 GB
-
49 6down 6Qwen2.5-7B-Instruct
Qwen 7.6B
Downloads (30 days) 1,034,284
Likes 298
F16 16.6 GB
-
50 5up 5Simplicio-27B
wesleysimplicio 27.3B
Downloads (30 days) 1,008,768
Likes 3
Q8_0 30.8 GB
Estimated Memory at 8K: Estimated with a formula. Actual usage depends on your setup and settings.
About this ranking
- We count the download numbers (last 30 days) that Hugging Face publishes for each repository. When several sources publish the same model, we add up the sources we list.
- Only models that generate text (Hugging Face task such as "text generation" or "image-text-to-text") are ranked. Embedding, object detection, speech recognition and similar models are not included.
- Only sources whose numbers were refreshed from Hugging Face within the last 2 days are added (sources that dropped out of the top of the list, whose numbers are stale, are not counted).
- This ranks how much models are used, not their performance or quality. This site does not assign its own performance scores.
- Hugging Face counts every request to a GGUF file (including HEAD) as a download. A split file may be counted once per part for a single download.
- Downloads that do not go through Hugging Face, such as pulls from the official Ollama library (ollama.com), are not included.
Source: Hugging Face Hub API (fetched daily around 04:20 JST)