Can gpt-oss Safeguard 20B run on the H200?
Yes. At 8K context the best fit is F16, using about 13.8 GB of the 134 GB usable on the H200. Estimated speed ceiling: ~1563 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) • | 11.6 GB Fits | 11.7 GB Fits | 12.2 GB Fits | 14.5 GB Fits | 1880 |
| 3-bit (Q3_K_M) • | 11.6 GB Fits | 11.7 GB Fits | 12.3 GB Fits | 14.5 GB Fits | 1873 |
| 4-bit (Q4_K_M) • | 11.7 GB Fits | 11.8 GB Fits | 12.4 GB Fits | 14.6 GB Fits | 1854 |
| 5-bit (Q5_K_M) • | 11.8 GB Fits | 11.9 GB Fits | 12.4 GB Fits | 14.7 GB Fits | 1840 |
| 6-bit (Q6_K) • | 12.1 GB Fits | 12.2 GB Fits | 12.8 GB Fits | 15.0 GB Fits | 1790 |
| 8-bit (Q8_0) • | 12.2 GB Fits | 12.3 GB Fits | 12.8 GB Fits | 15.1 GB Fits | 1780 |
| 16-bit (F16/BF16) • | 13.7 GB Fits | 13.8 GB Fits | 14.4 GB Fits | 16.6 GB Fits | 1563 |
• real GGUF file size from unsloth/gpt-oss-safeguard-20b-GGUF. Speed is a bandwidth ceiling (4800 GB/s), not a benchmark. Methodology.
What fits instead
- Cheapest listed GPU that runs gpt-oss Safeguard 20B: Arc A770 16GB (16 GB, launch MSRP $349) at Q8_0, fully in VRAM at 8K context.
- Largest popular model that fits the H200: Qwen3.8 Flash Next (180B, MoE) at Q5_K_M, about 121 GB at 8K context.
"Runs" and "fits" mean the whole model in GPU memory at 8K context (4-bit or better for the model pick). Prices are US launch MSRPs, not today's street prices. "Popular" means the 60 most-downloaded models we track.
More
gpt-oss Safeguard 20B on every GPU and Mac · Every model that runs on the H200 · Other models that fit 64 GB VRAM · Try other settings in the calculator
Guides: Best GPU for local LLMs by budget · Can I run gpt-oss, Llama 70B or Qwen3? · Mac vs PC for local LLMs
Data: Hugging Face Hub (openai/gpt-oss-safeguard-20b), NVIDIA specs. Model data last checked 2026-11-17; refreshed hourly.