circle-check\ Introducing Unsloth Studio: a new web UI for local AI\ \ 🦥chevron-right close
Unsloth LLMs directory for all our Dynamicarrow-up-right GGUF, 4-bit, 16-bit models on Hugging Face.
• GGUF + 4-bit
• Instruct 16-bit
• Base 4 & 16-bit
• FP8
Qwen DeepSeek Gemma Llama Mistral GLM
GGUFs let you run models in tools like Unsloth Studio ✨, Ollama and llama.cpp. Instruct (4-bit) safetensors can be used for inference or fine-tuning via Unsloth.
hashtag New & recommended models:
Model
Variant
GGUF
Instruct (4-bit)
27B
—
35B-A3B
—
26B-A4B
—
31B
E4B
E2B
Kimi
—
Nano-Omni-30B-A3B
—
35B-A3B
—
27B
—
122B-A10B
—
0.8B
—
2B
—
4B
—
9B
—
397B-A17B
—
Qwen3
—
NVIDIA Nemotron 3
—
GLM
—
—
Kimi
—
120B
20B
MiniMax
—
NVIDIA Nemotron 3
30B
—
2512
—
Edit-2511
—
3B
Instructarrow-up-right • Reasoningarrow-up-right
Instructarrow-up-right • Reasoningarrow-up-right
8B
Instructarrow-up-right • Reasoningarrow-up-right
Instructarrow-up-right • Reasoningarrow-up-right
14B
Instructarrow-up-right • Reasoningarrow-up-right
Instructarrow-up-right • Reasoningarrow-up-right
24B
—
123B
—
Mistral Large 3
675B
80B-A3B-Instruct
80B-A3B-Thinking
—
2B-Instruct
2B-Thinking
4B-Instruct
4B-Thinking
8B-Instruct
8B-Thinking
30B-A3B-Instruct
—
30B-A3B-Thinking
—
32B-Instruct
32B-Thinking
235B-A22B-Instruct
—
235B-A22B-Thinking
—
30B-A3B-Instruct
—
30B-A3B-Thinking
—
235B-A22B-Instruct
—
30B-A3B
—
4.7
—
4.6V-Flash
—
Terminus
—
V3.1
—
Granite-4.0
H-Small
Kimi-K2
Thinking
—
0905
—
hashtag DeepSeek models:
Model
Variant
GGUF
Instruct (4-bit)
DeepSeek-V3.1
Terminus
V3.1
DeepSeek-V3
V3-0324
—
V3
—
DeepSeek-R1
R1-0528
—
R1-0528-Qwen3-8B
R1
—
R1 Zero
—
Distill Llama 3 8 B
Distill Llama 3.3 70 B
Distill Qwen 2.5 1.5 B
Distill Qwen 2.5 7 B
Distill Qwen 2.5 14 B
Distill Qwen 2.5 32 B
hashtag Llama models:
Model
Variant
GGUF
Instruct (4-bit)
Llama 4
Scout 17 B-16 E
Maverick 17 B-128 E
—
Llama 3.3
70 B
Llama 3.2
1 B
3 B
11 B Vision
—
90 B Vision
—
Llama 3.1
8 B
70 B
—
405 B
—
Llama 3
8 B
—
70 B
—
Llama 2
7 B
—
13 B
—
CodeLlama
7 B
—
13 B
—
34 B
—
hashtag Gemma models:
Model
Variant
GGUF
Instruct (4-bit)
Gemma 4
E2B
E4B
26B-A4B
—
31B
FunctionGemma
270M
—
Gemma 3n
E2B
E4B
Gemma 3
270M
1 B
4 B
12 B
27 B
MedGemma
4 B (vision)
27 B (vision)
Gemma 2
2 B
9 B
—
27 B
—
hashtag Qwen models:
Model
Variant
GGUF
Instruct (4-bit)
27B
—
35B-A3B
—
35B-A3B
—
27B
—
122B-A10B
—
0.8B
—
2B
—
4B
—
9B
—
397B-A17B
—
Qwen3
—
2512
—
Edit-2511
—
2B-Instruct
2B-Thinking
4B-Instruct
4B-Thinking
8B-Instruct
8B-Thinking
Qwen3-Coder
30B-A3B
—
480B-A35B
—
30B-A3B-Instruct
—
30B-A3B-Thinking
—
235B-A22B-Thinking
—
235B-A22B-Instruct
—
Qwen 3
0.6 B
1.7 B
4 B
8 B
14 B
30 B-A3B
32 B
235 B-A22B
—
Qwen 2.5 Omni
3 B
—
7 B
—
Qwen 2.5 VL
3 B
7 B
32 B
72 B
Qwen 2.5
0.5 B
—
1.5 B
—
3 B
—
7 B
—
14 B
—
32 B
—
72 B
—
Qwen 2.5 Coder (128 K)
0.5 B
1.5 B
3 B
7 B
14 B
32 B
QwQ
32 B
QVQ (preview)
72 B
—
Qwen 2 (chat)
1.5 B
—
7 B
—
72 B
—
Qwen 2 VL
2 B
—
7 B
—
72 B
—
hashtag GLM models:
Model
Variant
GGUF
Instruct (4-bit)
GLM
—
—
4.6V-Flash
—
4.6
—
4.5-Air
—
hashtag Mistral models:
Model
Variant
GGUF
Instruct (4-bit)
Magistral
Small (2506)
Small (2509)
Small (2507)
Mistral Small
3.2-24 B (2506)
3.1-24 B (2503)
3-24 B (2501)
2409-22 B
—
Devstral
Small-24 B (2507)
Small-24 B (2505)
Pixtral
12 B (2409)
—
Mistral NeMo
12 B (2407)
Mistral Large
2407
—
Mistral 7 B
v0.3
—
v0.2
—
Mixtral
8 × 7 B
—
hashtag Phi models:
Model
Variant
GGUF
Instruct (4-bit)
Phi-4
Reasoning-plus
Reasoning
Mini-Reasoning
Phi-4 (instruct)
mini (instruct)
Phi-3.5
mini
—
Phi-3
mini
—
medium
—
hashtag Other (GLM, Orpheus, Smol, Llava etc.) models:
Model
Variant
GGUF
Instruct (4-bit)
GLM
4.5-Air
—
4.5
—
4-32B-0414
—
Grok 2
270B
—
Baidu-ERNIE
4.5-21B-A3B-Thinking
—
Hunyuan
A13B
—
Orpheus
0.1-ft (3B)
LLava
1.5 (7 B)
—
1.6 Mistral (7 B)
—
TinyLlama
Chat
—
SmolLM 2
135 M
360 M
1.7 B
Zephyr-SFT
7 B
—
Yi
6 B (v1.5)
—
6 B (v1.0)
—
34 B (chat)
—
34 B (base)
—
16-bit and 8-bit Instruct models are used for inference or fine-tuning in Unsloth Studio:
New models:
Model
Variant
Instruct (16-bit)
gpt-oss (new)
20b
120b
Gemma 3n
E2B
E4B
DeepSeek-R1-0528
R1-0528-Qwen3-8B
R1-0528
Mistral
Small 3.2 24B (2506)
Small 3.1 24B (2503)
Small 3.0 24B (2501)
Magistral Small (2506)
Qwen 3
0.6 B
1.7 B
4 B
8 B
14 B
30B-A3B
32 B
235B-A22B
Llama 4
Scout 17B-16E
Maverick 17B-128E
Qwen 2.5 Omni
3 B
7 B
Phi-4
Reasoning-plus
Reasoning
DeepSeek models
Model
Variant
Instruct (16-bit)
DeepSeek-V3
V3-0324
V3
DeepSeek-R1
R1-0528
R1-0528-Qwen3-8B
R1
R1 Zero
Distill Llama 3 8B
Distill Llama 3.3 70B
Distill Qwen 2.5 1.5B
Distill Qwen 2.5 7B
Distill Qwen 2.5 14B
Distill Qwen 2.5 32B
Llama models
Family
Variant
Instruct (16-bit)
Llama 4
Scout 17B-16E
Maverick 17B-128E
Llama 3.3
70 B
Llama 3.2
1 B
3 B
11 B Vision
90 B Vision
Llama 3.1
8 B
70 B
405 B
Llama 3
8 B
70 B
Llama 2
7 B
Gemma models:
Model
Variant
Instruct (16-bit)
Gemma 3n
E2B
E4B
Gemma 3
1 B
4 B
12 B
27 B
Gemma 2
2 B
9 B
27 B
Qwen models:
Family
Variant
Instruct (16-bit)
Qwen 3
0.6 B
1.7 B
4 B
8 B
14 B
30B-A3B
32 B
235B-A22B
Qwen 2.5 Omni
3 B
7 B
Qwen 2.5 VL
3 B
7 B
32 B
72 B
Qwen 2.5
0.5 B
1.5 B
3 B
7 B
14 B
32 B
72 B
Qwen 2.5 Coder 128 K
0.5 B
1.5 B
3 B
7 B
14 B
32 B
QwQ
32 B
QVQ (preview)
72 B
—
Qwen 2 (Chat)
1.5 B
7 B
72 B
Qwen 2 VL
2 B
7 B
72 B
Mistral models:
Model
Variant
Instruct (16-bit)
Mistral
Small 2409-22B
Mistral
Large 2407
Mistral
7B v0.3
Mistral
7B v0.2
Pixtral
12B 2409
Mixtral
8×7B
Mistral NeMo
12B 2407
Devstral
Small 2505
Phi models:
Model
Variant
Instruct (16-bit)
Phi-4
Reasoning-plus
Reasoning
Phi-4 (core)
Mini-Reasoning
Mini
Phi-3.5
Mini
Phi-3
Mini
Medium
Text-to-Speech (TTS) models:
Model
Instruct (16-bit)
Orpheus-3B (v0.1 ft)
Orpheus-3B (v0.1 pt)
Sesame-CSM 1B
Whisper Large V3 (STT)
Llasa-TTS 1B
Spark-TTS 0.5B
Oute-TTS 1B
Base models are usually used for fine-tuning purposes:
New models:
Model
Variant
Base (16-bit)
Base (4-bit)
Gemma 3n
E2B
E4B
Qwen 3
0.6 B
1.7 B
4 B
8 B
14 B
30B-A3B
Llama 4
Scout 17B 16E
Maverick 17B 128E
—
Llama models:
Model
Variant
Base (16-bit)
Base (4-bit)
Llama 4
Scout 17B 16E
—
Maverick 17B 128E
—
Llama 3.3
70 B
—
Llama 3.2
1 B
—
3 B
—
11 B Vision
—
90 B Vision
—
Llama 3.1
8 B
—
70 B
—
Llama 3
8 B
Llama 2
7 B
13 B
Qwen models:
Model
Variant
Base (16-bit)
Base (4-bit)
Qwen 3
0.6 B
1.7 B
4 B
8 B
14 B
30B-A3B
Qwen 2.5
0.5 B
1.5 B
3 B
7 B
14 B
32 B
72 B
Qwen 2
1.5 B
7 B
Llama models:
Model
Variant
Base (16-bit)
Base (4-bit)
Llama 4
Scout 17B 16E
—
Maverick 17B 128E
—
Llama 3.3
70 B
—
Llama 3.2
1 B
—
3 B
—
11 B Vision
—
90 B Vision
—
Llama 3.1
8 B
—
70 B
—
Llama 3
8 B
Llama 2
7 B
13 B
Gemma models
Model
Variant
Base (16-bit)
Base (4-bit)
Gemma 3
1 B
4 B
12 B
27 B
Gemma 2
2 B
—
9 B
—
27 B
—
Mistral models:
Model
Variant
Base (16-bit)
Base (4-bit)
Mistral
Small 24B 2501
—
NeMo 12B 2407
—
7B v0.3
7B v0.2
Pixtral 12B 2409
—
Other (TTS, TinyLlama) models:
Model
Variant
Base (16-bit)
Base (4-bit)
TinyLlama
1.1 B (Base)
Orpheus-3b
0.1-pretrained
You can use our FP8 uploads for training or serving/deployment.
FP8 Dynamic offers slightly faster training and lower VRAM usage than FP8 Block, but with a small trade-off in accuracy.
Model
Variant
FP8 (Dynamic / Block)
Qwen3
Coder-Next
Dynamicarrow-up-right · Blockarrow-up-right
GLM
4.7-Flash
Llama 3.3
70B Instruct
Dynamicarrow-up-right · Blockarrow-up-right
Llama 3.2
1B Base
Dynamicarrow-up-right · Blockarrow-up-right
1B Instruct
Dynamicarrow-up-right · Blockarrow-up-right
3B Base
Dynamicarrow-up-right · Blockarrow-up-right
3B Instruct
Dynamicarrow-up-right · Blockarrow-up-right
Llama 3.1
8B Base
Dynamicarrow-up-right · Blockarrow-up-right
8B Instruct
Dynamicarrow-up-right · Blockarrow-up-right
70B Base
Dynamicarrow-up-right · Blockarrow-up-right
Qwen3
0.6B
1.7B
4B
8B
14B
32B
235B-A22B
Qwen3 (2507)
4B Instruct
4B Thinking
30B-A3B Instruct
30B-A3B Thinking
235B-A22B Instruct
235B-A22B Thinking
Qwen3-VL
4B Instruct
4B Thinking
8B Instruct
8B Thinking
Qwen3-Coder
480B-A35B Instruct
Granite 4.0
h-tiny
h-small
Magistral Small
2509
FP8 Dynamicarrow-up-right · FP8 torchaoarrow-up-right
Mistral Small 3.2
24B Instruct-2506
Gemma 3
270M-it torchao 270m — FP8arrow-up-right 1B — FP8arrow-up-right 4B — FP8arrow-up-right 12B — FP8arrow-up-right 27B — FP8arrow-up-right
PreviousUnsloth Notebookschevron-left NextInstallationchevron-right
Last updated 11 hours ago
Was this helpful?
This site uses cookies to deliver its service and to analyze traffic. By browsing this site, you accept the privacy policy.
close
AcceptReject