Can Llama Guard 4 12B run on the H100 PCIe?
Yes. At 8K context the best fit is F16, using about 24.6 GB of the 76.0 GB usable on the H100 PCIe. Estimated speed ceiling: ~83 tokens/sec.
Memory by quant and context
| Quant | 4K ctx | 8K ctx | 32K ctx | 128K ctx | Max tok/s (est., 8K) |
|---|---|---|---|---|---|
| 2-bit (Q2_K) | 5.7 GB Fits | 6.5 GB Fits | 11.0 GB Fits | 29.0 GB Fits | 444 |
| 3-bit (Q3_K_M) | 7.0 GB Fits | 7.7 GB Fits | 12.2 GB Fits | 30.2 GB Fits | 341 |
| 4-bit (Q4_K_M) | 8.4 GB Fits | 9.1 GB Fits | 13.6 GB Fits | 31.6 GB Fits | 273 |
| 5-bit (Q5_K_M) | 9.5 GB Fits | 10.2 GB Fits | 14.7 GB Fits | 32.7 GB Fits | 234 |
| 6-bit (Q6_K) | 10.7 GB Fits | 11.5 GB Fits | 16.0 GB Fits | 34.0 GB Fits | 203 |
| 8-bit (Q8_0) | 13.4 GB Fits | 14.2 GB Fits | 18.7 GB Fits | 36.7 GB Fits | 157 |
| 16-bit (F16/BF16) | 23.9 GB Fits | 24.6 GB Fits | 29.1 GB Fits | 47.1 GB Fits | 83 |
• real GGUF file size. Speed is a bandwidth ceiling (2000 GB/s), not a benchmark. Methodology.
More
Llama Guard 4 12B on every GPU and Mac · Every model that runs on the H100 PCIe · Other models that fit 64 GB VRAM · Try other settings in the calculator
Data: Hugging Face Hub (meta-llama/Llama-Guard-4-12B), NVIDIA specs. Data as of 2026-11-17.