Can GeForce RTX 4090 run Llama 3.2 3B?

✅ Yes, runs great!

Fits in GPU memory (~208 tok/s)

Memory Needed
5.0 GB
Q8_0, 8K context
Est. Speed
~208 tok/s
Generation speed
GPU VRAM
24 GB
GDDR6X

💡 Recommended

For best quality that fits: Q8_0 quantization

All Quantizations

Q8_0
RECOMMENDED
Runs great
Memory
5.0 GB
Speed
208 tok/s
Bits/weight
8.5
Quality
Highest
Q6_K
Runs great
Memory
4.2 GB
Speed
267 tok/s
Bits/weight
6.6
Quality
High
Q5_K_M
Runs great
Memory
3.9 GB
Speed
309 tok/s
Bits/weight
5.7
Quality
High
Q4_K_M
Runs great
Memory
3.6 GB
Speed
364 tok/s
Bits/weight
4.85
Quality
Good
Q3_K_M
Runs great
Memory
3.2 GB
Speed
452 tok/s
Bits/weight
3.9
Quality
Lower
Q2_K
Runs great
Memory
2.9 GB
Speed
588 tok/s
Bits/weight
3
Quality
Lower

Run with Ollama

ollama run llama3.2:3b-q8_0

This will download and run the Q8_0 quantization

View all GPUs for Llama 3.2 3BView all models for GeForce RTX 4090