ASUS PX13 Strix Halo runs 120B LLMs

// 45d agoBENCHMARK RESULT

ASUS PX13 Strix Halo runs 120B LLMs

Benchmarks show the 256-bit memory bus on AMD's new APU enables 120B parameter models to run at 15 tok/s on a 13-inch chassis. A significant leap for local AI workstation portability that finally bridges the gap with multi-GPU desktops.

// ANALYSIS

Strix Halo is the hardware breakthrough local AI enthusiasts have waited for, bringing desktop-class unified memory bandwidth to a portable form factor. The 128GB LPDDR5x-8000 configuration provides approximately 215 GB/s bandwidth, while Vulkan (RADV) emerges as the superior backend for consumer Linux over ROCm. Performance scaling at 40W "Power Saver" mode keeps the 13-inch chassis quiet during inference, and the amd-strix-halo-toolboxes project simplifies the complex dependency chain via containerization.

// TAGS

llmbenchmarkgpuinferenceopen-sourceasus-proart-px13strix-halo

DISCOVERED

45d ago

2026-04-12

PUBLISHED

46d ago

2026-04-12

RELEVANCE

8/ 10

AUTHOR

Willing-Toe1942

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

MODEL3h ago

Anthropic drops Opus 4.8 for Claude Code

Anthropic has released Opus 4.8, integrating the new model into Claude Code with high-effort defaults for complex coding tasks. The update boosts SWE-bench Pro scores to 69.2% and drastically reduces unremarked flaws in generated code.

VIDEO3h ago

Google AI animates cardboard TPUs for I/O 2026

Google AI partners with director Laurie Rowan and Nexus Studios to create a promotional short film for Google I/O 2026. The project leverages AI models to animate physical materials like cardboard and markers into characters representing Tensor Processing Units.

MODEL3h ago

Claude Opus 4.8 drops with extended agentic autonomy

Anthropic has released Claude Opus 4.8, bringing improvements to agentic skills, reasoning, and coding capabilities at the exact same price. The update introduces sharper judgment, increased honesty about its task progress, and the ability to operate autonomously for much longer periods.