Can GeForce RTX 3060 12GB run Qwen 2.5 32B?

⚠️ Runs slowly (CPU offload needed)

Needs CPU offload (21.6 GB total). Estimated 3 tok/s

Memory Needed
21.6 GB
Q4_K_M, 8K context
Est. Speed
~3 tok/s
Generation speed
GPU VRAM
12 GB
GDDR6

💡 Recommended

For best quality that fits: Q4_K_M quantization

All Quantizations

Q8_0
Won't run
Memory
37.3 GB
Speed
N/A
Bits/weight
8.5
Quality
Highest
Q6_K
Won't run
Memory
29.6 GB
Speed
N/A
Bits/weight
6.6
Quality
High
Q5_K_M
Won't run
Memory
24.3 GB
Speed
N/A
Bits/weight
5.7
Quality
High
Q4_K_M
RECOMMENDED
Runs slowly
Memory
21.6 GB
Speed
3 tok/s
Bits/weight
4.85
Quality
Good
Q3_K_M
Runs slowly
Memory
18.1 GB
Speed
3 tok/s
Bits/weight
3.9
Quality
Lower
Q2_K
Runs slowly
Memory
14.8 GB
Speed
4 tok/s
Bits/weight
3
Quality
Lower

Run with Ollama

ollama run qwen2.5:32b

This will download and run the Q4_K_M quantization

View all GPUs for Qwen 2.5 32BView all models for GeForce RTX 3060 12GB