▣~/runthismodel▍
daemon okbuild 5a3c91d00:00:00Z
Back to RunThisModel

AI Models You Can Run with 48GB VRAM

142 models compatible across 9 categories

Compatible Models
142
Largest Model
Magnum v4 72B (72B)
Categories
9
4GB6GB8GB10GB12GB16GB20GB24GB32GB48GB64GB80GB128GB192GB
💬

Chat / LLM (72 models)

Magnum v4 72B

Anthracite

72B
44.7GB VRAMQ4_K_M
85%

Llama 3.1 70B Instruct

Meta

70B
40.1GB VRAMQ4_K_M
85%

Euryale L3.3 70B v2.3

Sao10K

70B
40.1GB VRAMQ4_K_M
85%

Llama 3.1 70B (lorablated)

mlabonne

70B
40.1GB VRAMQ4_K_M
85%

Mixtral 8x7B Instruct

Mistral AI

46.7B
30.5GB VRAMQ5_K_M
92%

Phi-3.5 MoE

Microsoft

41.9B
24.1GB VRAMQ4_K_M
85%

Qwen 2.5 32B

Alibaba

32B
19.0GB VRAMQ4_K_M
85%

Skyfall 31B v4.2

TheDrummer

31B
18.2GB VRAMQ4_K_M
85%

Qwen3 30B-A3B

Alibaba

30.5B
36.0GB VRAMQ8_0
98%

Gemma 3 27B

Google

27B
15.9GB VRAMQ4_K_M
85%

Dolphin Mistral 24B (Venice Edition)

Cognitive Computations

24B
25.9GB VRAMQ8_0
98%

Dolphin 3.0 R1 Mistral 24B

Cognitive Computations

24B
13.8GB VRAMQ4_K_M
85%

Cydonia 24B v4.3

TheDrummer

24B
13.8GB VRAMQ4_K_M
85%

Mistral Small 22B

Mistral AI

22B
12.9GB VRAMQ4_K_M
85%

Magnum v4 22B

Anthracite

22B
44.5GB VRAMBF16
100%

DeepSeek MoE 16B

DeepSeek

16.4B
11.0GB VRAMQ4_K_M
85%

Rocinante XL 16B v1

TheDrummer

16B
32.5GB VRAMBF16
100%

Phi-4

Microsoft

14B
15.0GB VRAMQ8_0
98%

Qwen 2.5 14B

Alibaba

14B
15.1GB VRAMQ8_0
98%

Gemma 3 12B

Google

12B
12.2GB VRAMQ8_0
98%

Mistral Nemo 12B

Mistral AI

12B
12.6GB VRAMQ8_0
98%

Magnum v4 12B

Anthracite

12B
24.5GB VRAMBF16
100%

Rocinante 12B v1.1

TheDrummer

12B
24.5GB VRAMBF16
100%

Mistral Nemo Base 12B

Mistral AI

12B
24.5GB VRAMBF16
100%

Solar 10.7B

Upstage

10.7B
11.1GB VRAMQ8_0
98%

Falcon 3 10B

TII

10B
10.7GB VRAMQ8_0
98%

Gemma 2 9B Instruct

Google

9.2B
9.7GB VRAMQ8_0
98%

Yi 1.5 9B Chat

01.AI

9B
9.2GB VRAMQ8_0
98%

Gemma 3 MoE 9B

Google

9B
7.0GB VRAMQ4_K_M
85%

DeepSeek R1 Distill 8B

DeepSeek

8B
8.4GB VRAMQ8_0
98%

Llama 3.1 8B Instruct

Meta

8B
17.0GB VRAMFP16
100%

Granite 3.3 8B

IBM

8B
8.6GB VRAMQ8_0
98%

Dolphin 3.0 Llama 3.1 8B

Cognitive Computations

8B
16.5GB VRAMBF16
100%

NeuralDaredevil 8B (abliterated)

mlabonne

8B
16.5GB VRAMBF16
100%

Llama 3.1 8B Instruct (abliterated)

mlabonne

8B
16.5GB VRAMBF16
100%

Stheno L3 8B v3.2

Sao10K

8B
16.5GB VRAMBF16
100%

Qwen3 8B Base

Alibaba

8B
16.5GB VRAMBF16
100%

EXAONE 3.5 7.8B

LG AI

7.8B
8.2GB VRAMQ8_0
98%

InternLM 2.5 7B

Shanghai AI Lab

7.7B
8.2GB VRAMQ8_0
98%

Qwen 2.5 7B Instruct

Alibaba

7.6B
9.0GB VRAMQ8_0
98%

Mistral 7B Instruct v0.3

Mistral AI

7.3B
15.5GB VRAMFP16
100%

Falcon 3 7B

TII

7B
8.3GB VRAMQ8_0
98%

OLMo 2 7B

Allen AI

7B
7.7GB VRAMQ8_0
98%

OpenChat 3.5 7B

OpenChat

7B
7.7GB VRAMQ8_0
98%

OLMoE 1B-7B

AI2

6.9B
7.3GB VRAMQ8_0
98%

Yi 1.5 6B Chat

01.AI

6B
6.5GB VRAMQ8_0
98%

Gemma 3 4B

Google

4B
4.3GB VRAMQ8_0
98%

Nemotron Mini 4B

NVIDIA

4B
4.7GB VRAMQ8_0
98%

Danube 3 4B

H2O.ai

4B
4.4GB VRAMQ8_0
98%

Phi-3.5 Mini 3.8B

Microsoft

3.8B
4.3GB VRAMQ8_0
98%

Phi-4 Mini 3.8B

Microsoft

3.8B
4.3GB VRAMQ8_0
98%

Granite 3.0 3B-A800M

IBM

3.4B
2.4GB VRAMQ4_K_M
85%

Llama 3.2 3B Instruct

Meta

3.2B
3.7GB VRAMQ8_0
98%

Qwen 2.5 3B

Alibaba

3B
3.9GB VRAMQ8_0
98%

Falcon 3 3B

TII

3B
3.8GB VRAMQ8_0
98%

StableLM Zephyr 3B

Stability AI

3B
3.3GB VRAMQ8_0
98%

Rocket 3B

Pansophic

3B
3.3GB VRAMQ8_0
98%

Gemma 2 2B

Google

2.6B
3.1GB VRAMQ8_0
98%

EXAONE 3.5 2.4B

LG AI

2.4B
3.1GB VRAMQ8_0
98%

Granite 3.3 2B

IBM

2B
3.0GB VRAMQ8_0
98%

SmolLM2 1.7B

HuggingFace

1.7B
2.2GB VRAMQ8_0
98%

Qwen 2.5 1.5B

Alibaba

1.5B
2.3GB VRAMQ8_0
98%

DeepSeek R1 Distill 1.5B

DeepSeek

1.5B
2.3GB VRAMQ8_0
98%

Granite 3.0 1B-A400M

IBM

1.3B
1.3GB VRAMQ4_K_M
85%

Llama 3.2 1B Instruct

Meta

1.24B
2.8GB VRAMFP16
100%

TinyLlama 1.1B

TinyLlama

1.1B
1.6GB VRAMQ8_0
98%

Gemma 3 1B

Google

1B
1.5GB VRAMQ8_0
98%

Falcon 3 1B

TII

1B
2.2GB VRAMQ8_0
98%

Qwen 2.5 0.5B

Alibaba

0.5B
1.1GB VRAMQ8_0
98%

Danube 3 500M

H2O.ai

0.5B
1.0GB VRAMQ8_0
98%

SmolLM2 360M

HuggingFace

0.36B
0.9GB VRAMQ8_0
98%

SmolLM2 135M

HuggingFace

0.135B
0.8GB VRAMFP16
100%
💻

Coding (17 models)

Codestral 22B (abliterated)

failspy

22B
44.5GB VRAMBF16
100%

Qwen 2.5 Coder 14B

Alibaba

14B
15.1GB VRAMQ8_0
98%

Code Llama 13B Instruct

Meta

13B
7.8GB VRAMQ4_K_M
85%

Yi Coder 9B

01.AI

9B
9.2GB VRAMQ8_0
98%

CodeGemma 7B

Google

8.5B
8.9GB VRAMQ8_0
98%

Qwen 2.5 Coder 7B

Alibaba

7.6B
8.0GB VRAMQ8_0
98%

StarCoder2 7B

BigCode

7B
7.6GB VRAMQ8_0
98%

Code Llama 7B

Meta

7B
7.2GB VRAMQ8_0
98%

DeepSeek Coder 6.7B

DeepSeek

6.7B
7.2GB VRAMQ8_0
98%

Qwen 2.5 Coder 3B

Alibaba

3B
3.9GB VRAMQ8_0
98%

StarCoder2 3B

BigCode

3B
3.5GB VRAMQ8_0
98%

Stable Code 3B

Stability AI

3B
3.3GB VRAMQ8_0
98%

CodeGemma 2B

Google

2B
3.0GB VRAMQ8_0
98%

Qwen 2.5 Coder 1.5B

Alibaba

1.5B
2.3GB VRAMQ8_0
98%

Yi Coder 1.5B

01.AI

1.5B
2.0GB VRAMQ8_0
98%

DeepSeek Coder 1.3B

DeepSeek

1.3B
1.8GB VRAMQ8_0
98%

Qwen 2.5 Coder 0.5B

Alibaba

0.5B
1.1GB VRAMQ8_0
98%
👁

Multimodal (6 models)

LLaVA 1.6 7B

LLaVA

7B
8.5GB VRAMQ8_0
98%

Phi-3.5 Vision

Microsoft

4.2B
3.2GB VRAMQ4_K_M
85%

PaliGemma 3B

Google

3B
2.5GB VRAMQ4_K_M
85%

Qwen2-VL 2B

Alibaba

2.2B
2.0GB VRAMQ8_0
98%

MiniCPM-V 2.6

OpenBMB

2B
3.0GB VRAMQ8_0
98%

Moondream 2

Moondream

1.8B
1.5GB VRAMQ4_K_M
85%
🎨

Image Generation (9 models)

FLUX.1 Schnell (GGUF)

Black Forest Labs

12B
14.0GB VRAMQ5_0
90%

FLUX.1 Dev (GGUF)

Black Forest Labs

12B
14.0GB VRAMQ5_0
100%

Stable Diffusion XL (CoreML)

Stability AI

3.5B
3.3GB VRAMCoreML
100%

SDXL Turbo (GGUF)

Stability AI

3.5B
5.0GB VRAMQ5_0
85%

Stable Diffusion 3 Medium (GGUF)

Stability AI

2.5B
9.2GB VRAMQ8_0
95%

Stable Diffusion 2.1 Base (CoreML)

Stability AI / Apple

0.86B
1.6GB VRAMCoreML-Palettized
85%

Stable Diffusion 1.5 (CoreML)

Runway

0.86B
2.5GB VRAMCoreML-Palettized
90%

Stable Diffusion 1.5 (GGUF)

Runway / GPUStack

0.86B
2.3GB VRAMQ8_0
95%

Stable Diffusion 2.1 (GGUF)

Stability AI

0.86B
2.7GB VRAMQ8_0
95%
🎤

Speech Recognition (9 models)

Whisper Large v3

OpenAI

1.55B
3.4GB VRAMQ8_0
98%

Whisper Large v3 Turbo

OpenAI

0.81B
2.0GB VRAMQ8_0
95%

Whisper Medium

OpenAI

0.77B
1.9GB VRAMQ8_0
92%

Distil-Whisper Large v3

HuggingFace

0.76B
1.9GB VRAMQ8_0
96%

Whisper Small

OpenAI

0.24B
0.9GB VRAMQ8_0
85%

Whisper Base

OpenAI

0.074B
0.3GB VRAMQ8_0
80%

Whisper Base English

OpenAI

0.074B
0.3GB VRAMQ8_0
82%

Whisper Tiny English (Quantized)

OpenAI

0.039B
0.1GB VRAMQ5_1
65%

Whisper Tiny

OpenAI

0.039B
0.2GB VRAMQ8_0
70%
🔊

Text to Speech (14 models)

Kokoro 82M TTS

Kokoro

0.082B
0.6GB VRAMONNX-Q8F16
95%

Piper TTS - Amy (English)

Rhasspy

0.02B
0.1GB VRAMONNX
85%

Piper TTS - Lessac (English)

Rhasspy

0.02B
0.1GB VRAMONNX
85%

Piper TTS - LibriTTS-R (English)

Rhasspy

0.02B
0.6GB VRAMONNX
80%

Piper TTS - Spanish (MLS)

Rhasspy

0.02B
0.1GB VRAMONNX
80%

Piper TTS - French (Siwis)

Rhasspy

0.02B
0.5GB VRAMONNX
80%

Piper TTS - German (Thorsten)

Rhasspy

0.02B
0.1GB VRAMONNX
80%

Piper TTS - Chinese (Huayan)

Rhasspy

0.02B
0.1GB VRAMONNX
80%

Piper TTS - Japanese (Kokoro)

Rhasspy

0.02B
0.1GB VRAMONNX
80%

Piper TTS - Korean

Rhasspy

0.02B
0.1GB VRAMONNX
80%

Piper TTS - Russian (Irina)

Rhasspy

0.02B
0.1GB VRAMONNX
80%

Piper TTS - Portuguese (Faber)

Rhasspy

0.02B
0.1GB VRAMONNX
80%

Piper TTS - Italian (Riccardo)

Rhasspy

0.02B
0.5GB VRAMONNX
80%

Piper TTS - Arabic (Kareem)

Rhasspy

0.02B
0.1GB VRAMONNX
80%
🎵

Audio Generation (3 models)

ACE-Step 1.5XL

ACE Studio

1.5B
8.0GB VRAMFP16
100%

Stable Audio Open

Stability AI

1B
6.0GB VRAMFP16
100%

MusicGen Small

Meta

0.3B
0.8GB VRAMONNX-Q4F16
80%
🧩

Embedding (5 models)

BGE Large EN v1.5

BAAI

0.335B
1.1GB VRAMFP16
100%

Nomic Embed Text v1.5

Nomic AI

0.137B
0.8GB VRAMFP16
100%

BGE Small EN v1.5

BAAI

0.033B
0.1GB VRAMQ8_0
90%

Snowflake Arctic Embed S

Snowflake

0.033B
0.1GB VRAMQ8_0
88%

all-MiniLM-L6-v2

Sentence Transformers

0.023B
0.1GB VRAMQ8_0
92%
🔄

Reranker (2 models)

BGE Reranker v2 M3

BAAI

0.568B
1.6GB VRAMFP16
98%

Jina Reranker Tiny EN

Jina AI

0.033B
0.1GB VRAMFP16
85%

Compatible GPUs (15 with 48GB+ VRAM)

Apple M4 Pro

48GBapple

NVIDIA RTX A6000

48GB$4,650nvidia

NVIDIA L40S

48GB$7,500nvidia

Apple M1 Max

64GBapple

NVIDIA A100 80GB

80GB$15,000nvidia

NVIDIA H100

80GB$30,000nvidia

Apple M2 Max

96GBapple

Apple M4 Max

128GBapple

Apple M3 Max

128GBapple

Apple M1 Ultra

128GBapple

AMD Instinct MI250X

128GB$10,000amd

Apple M4 Ultra

192GBapple

Apple M3 Ultra

192GBapple

Apple M2 Ultra

192GBapple

AMD Instinct MI300X

192GB$15,000amd

Frequently Asked Questions

What is the best AI model I can run with 48GB VRAM?
The largest model you can run with 48GB VRAM is Magnum v4 72B (72B parameters) using Q4_K_M quantization. There are 142 total compatible models.
Can I run Llama with 48GB VRAM?
Yes! You can run 10 Llama models with 48GB VRAM: Llama 3.1 70B Instruct (Q4_K_M), Llama 3.1 70B (lorablated) (Q4_K_M), Code Llama 13B Instruct (Q4_K_M), Llama 3.1 8B Instruct (FP16), Dolphin 3.0 Llama 3.1 8B (BF16), Llama 3.1 8B Instruct (abliterated) (BF16), Code Llama 7B (Q8_0), Llama 3.2 3B Instruct (Q8_0), Llama 3.2 1B Instruct (FP16), TinyLlama 1.1B (Q8_0).
What GPU has 48GB VRAM?
GPUs with 48GB or more VRAM include: Apple M4 Pro, NVIDIA RTX A6000, NVIDIA L40S, Apple M1 Max, NVIDIA A100 80GB, and 10 more.

Other VRAM Tiers

4GB VRAM6GB VRAM8GB VRAM10GB VRAM12GB VRAM16GB VRAM20GB VRAM24GB VRAM32GB VRAM64GB VRAM80GB VRAM128GB VRAM192GB VRAM
© runthismodel · 2026privacytermsdisclaimerchangelogembed badgerunpodvast.aihuggingfaceollamalm-studiomade for the people who actually read GGUF metadata
 ┌─┐                ╔══╗     ╔══╗
 │░│  RUN  THIS  M  ║▓▓║ DEL ║▓▓║
 └─┘                ╚══╝     ╚══╝