# Can I Run OpenAI gpt-oss 120B on AMD Strix Halo 128GB? Unified Memory Benchmark

> **Direct Compatibility Verdict**: Yes, OpenAI gpt-oss 120B fits completely inside Strix Halo's 96GB ROCm/Vulkan GTT unified memory pool at Q4_K_M (71.17 GB at 8k context, 77.79 GB at 128k context). Activating only 5.1B parameters per token sustains a fluid 32–38 tok/s locally with silent <85W power, whereas standard 48GB dual-3090 setups fail with devastating CPU RAM offloading.

## Key Memory Physics
| Metric | Value |
|---|---|
| Target AI Model | OpenAI gpt-oss 120B (MoE) |
| Target Hardware | AMD Ryzen AI Max+ 395 Strix Halo 128GB |
| Q4_K_M VRAM Required | 71.17 GB |
| Available Headroom | +24.83 GB |
| Estimated Generation Speed | ~35 tok/s |

## Hardware Physics & Bottleneck Analysis
Holding 116.8 billion parameters requires an enormous unified memory space. While traditional desktops split memory across two PCIe slots, AMD Strix Halo places 128GB of coherent LPDDR5X-8533 memory directly on the APU substrate (256 GB/s). ROCm and Vulkan can allocate up to 96GB to the integrated GPU, leaving 32GB for the operating system and IDE.

## Tier 1 Free Software Optimization ($0)
On Linux, configure the ROCm GTT memory ceiling via kernel boot parameter `amdgpu.gttsize=98304` to grant llama.cpp access to the full 96GB pool, and run `llama-server -fa on -ctk q8_0` for maximum prompt-processing throughput.

## Verified Hardware & Cloud Pathways
- Search Hardware on Amazon: https://www.amazon.com/s?k=AMD+Ryzen+AI+Max%2B+395+Strix+Halo+128GB&tag=sparecables-20 (rel="noopener sponsored")
- Rent on RunPod: https://runpod.io?ref=ynypzu94 (rel="noopener sponsored")
- Rent on Vast.ai: https://cloud.vast.ai/?ref_id=723130 (rel="noopener sponsored")

---
*Interactive Sizing Calculator available at: https://www.gpugarden.com/can-i-run/gpt-oss-120b-on-strix-halo-128gb/*