NVIDIA L40 meets the minimum VRAM requirement for Q4 inference of mlx-community/gpt-oss-20b-MXFP4-Q8. Review the quantization breakdown below to see how higher precision settings impact VRAM and throughput.
NVIDIA L40 can run mlx-community/gpt-oss-20b-MXFP4-Q8 with Q4 quantization. At approximately 97 tokens/second, you can expect Good speed - acceptable for interactive use.
You have 38GB headroom, which is sufficient for system overhead and smooth operation.
| Quantization | VRAM needed | VRAM available | Estimated speed | Verdict |
|---|---|---|---|---|
| Q4 | 10GB | 48GB | 96.56 tok/s | ✅ Fits comfortably |
| Q8 | 20GB | 48GB | 59.16 tok/s | ✅ Fits comfortably |
| FP16 | 41GB | 48GB | 37.95 tok/s | ✅ Fits comfortably |