NVIDIA L40 meets the minimum VRAM requirement for Q4 inference of codellama/CodeLlama-34b-hf. Review the quantization breakdown below to see how higher precision settings impact VRAM and throughput.
NVIDIA L40 can run codellama/CodeLlama-34b-hf with Q4 quantization. At approximately 54 tokens/second, you can expect Good speed - acceptable for interactive use.
You have 31GB headroom, which is sufficient for system overhead and smooth operation.
| Quantization | VRAM needed | VRAM available | Estimated speed | Verdict |
|---|---|---|---|---|
| Q4 | 17GB | 48GB | 53.73 tok/s | ✅ Fits comfortably |
| Q8 | 35GB | 48GB | 37.08 tok/s | ✅ Fits comfortably |
| FP16 | 70GB | 48GB | 23.83 tok/s | ❌ Not recommended |