RunsOnMy ModelsGPUs & MacsBy VRAMBest GPUCan I run…?API pricesVRAM calculatorNew

Can gpt-oss Safeguard 120B run on the RTX 6000 Ada?

Not fully in VRAM. Even at 2-bit with 8K context, gpt-oss Safeguard 120B needs about 59.3 GB, more than the 45.6 GB usable on the RTX 6000 Ada. It can still run with offload (slower): Q4_K_M with part of the model in system RAM, about 37–66 tokens/sec (estimate). Details and command.

Runs with offload (slower)

Runs with offload 4-bit (Q4_K_M) at 8K context: all attention layers plus the experts of 26 of 36 MoE layers on the RTX 6000 Ada (about 42.8 GB), and the experts of the other 10 layers in system RAM (about 15.9 GB). Needs a PC with at least 32 GB of system RAM. Estimated speed: 37–66 tokens/sec (usable for chat), versus a bandwidth ceiling of ~311 tok/s if it fit fully in VRAM.

llama-server -hf unsloth/gpt-oss-safeguard-120b-GGUF:Q4_K_M -c 8192 -ngl 999 --n-cpu-moe 10

Estimate, not a benchmark. The command downloads the GGUF from Hugging Face. The speed range spans dual-channel DDR4-3200 to DDR5-6000 system RAM; your CPU, RAM speed and llama.cpp version move it. If llama.cpp runs out of GPU memory, raise --n-cpu-moe by 1-2; if VRAM is left over, lower it for more speed. Recent llama.cpp builds can also pick the split automatically. How we estimate offload.

Memory by quant and context

Quant4K ctx8K ctx32K ctx128K ctxMax tok/s (est., 8K)
2-bit (Q2_K) •59.2 GB No59.3 GB No60.2 GB No63.6 GB No—
3-bit (Q3_K_M) •59.3 GB No59.4 GB No60.2 GB No63.6 GB No—
4-bit (Q4_K_M) •59.4 GB No59.5 GB No60.4 GB No63.7 GB No—
5-bit (Q5_K_M) •59.5 GB No59.6 GB No60.5 GB No63.9 GB No—
6-bit (Q6_K) •59.9 GB No60.0 GB No60.9 GB No64.2 GB No—
8-bit (Q8_0) •60.0 GB No60.1 GB No60.9 GB No64.3 GB No—
16-bit (F16/BF16)219 GB No219 GB No220 GB No223 GB No—

• real GGUF file size from unsloth/gpt-oss-safeguard-120b-GGUF. Speed is a bandwidth ceiling (960 GB/s), not a benchmark. Methodology.

Hardware that does fit it

A100 80GBH100 PCIeH100 SXMRTX PRO 6000 BlackwellMac M2 Max 96GBDGX Spark

What fits instead

"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 120B on every GPU and Mac · Every model that runs on the RTX 6000 Ada · Other models that fit 48 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-120b), NVIDIA specs. Model data last checked 2026-11-17; refreshed hourly.