NVIDIA L40S meets the minimum VRAM requirement for Q4 inference of baichuan-inc/Baichuan-M2-32B. Review the quantization breakdown below to see how higher precision settings impact VRAM and throughput.
NVIDIA L40S can run baichuan-inc/Baichuan-M2-32B with Q4 quantization. At approximately 58 tokens/second, you can expect Good speed - acceptable for interactive use.
You have 32GB headroom, which is sufficient for system overhead and smooth operation.
| Quantization | VRAM needed | VRAM available | Estimated speed | Verdict |
|---|---|---|---|---|
| Q4 | 16GB | 48GB | 57.92 tok/s | ✅ Fits comfortably |
| Q8 | 33GB | 48GB | 41.26 tok/s | ✅ Fits comfortably |
| FP16 | 66GB | 48GB | 22.05 tok/s | ❌ Not recommended |