NVIDIA L4 meets the minimum VRAM requirement for Q4 inference of meta-llama/Llama-3.2-1B. Review the quantization breakdown below to see how higher precision settings impact VRAM and throughput.
NVIDIA L4 can run meta-llama/Llama-3.2-1B with Q4 quantization. At approximately 75 tokens/second, you can expect Good speed - acceptable for interactive use.
You have 23GB headroom, which is sufficient for system overhead and smooth operation.
| Quantization | VRAM needed | VRAM available | Estimated speed | Verdict |
|---|---|---|---|---|
| Q4 | 1GB | 24GB | 74.81 tok/s | ✅ Fits comfortably |
| Q8 | 1GB | 24GB | 50.91 tok/s | ✅ Fits comfortably |
| FP16 | 2GB | 24GB | 25.67 tok/s | ✅ Fits comfortably |