NVIDIA L40S meets the minimum VRAM requirement for Q4 inference of openai/gpt-oss-20b. Review the quantization breakdown below to see how higher precision settings impact VRAM and throughput.
NVIDIA L40S can run openai/gpt-oss-20b with Q4 quantization. At approximately 83 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 | 82.73 tok/s | ✅ Fits comfortably |
| Q8 | 20GB | 48GB | 59.68 tok/s | ✅ Fits comfortably |
| FP16 | 41GB | 48GB | 38.04 tok/s | ✅ Fits comfortably |