Muse Glimmer on AMD Hardware

AMD and Meta partnered to optimize Muse Glimmer for Ryzen AI Max+ and Radeon GPUs. Here is the hardware that runs it best.

The AMD Partnership

Muse Glimmer was co-launched with AMD — a deliberate move by Meta to prove that local frontier AI does not require NVIDIA hardware. AMD contributed early access to Ryzen AI Max+ silicon and engineering support to optimize the ROCm inference path before public release.

The result: AMD hardware is a first-class citizen, not an afterthought. Where many open-weight models technically support ROCm but deliver poor performance, Muse Glimmer was specifically tuned for the AMD inference stack from day one.

For AMD users — whether you have a Radeon GPU or a Ryzen AI Max+ laptop — Muse Glimmer is the open-weight model built for your hardware.

Ryzen AI Max+ Advantage

Up to 128GB of unified memory means Muse Glimmer runs at near-full precision without quantization — the only consumer laptop platform that can do this. No external GPU required.

ROCm First-Class Support

ROCm 6.x enables full GPU acceleration on Radeon cards. Ollama ships with ROCm backend support — pull Muse Glimmer and it detects your AMD GPU automatically.

Official, Not Community

Unlike other models where AMD support is community-maintained, Muse Glimmer's AMD path is tested and supported by Meta and AMD jointly. Bug reports go to the same team that shipped the model.

AMD Hardware Options

Supported AMD configurations from the official partnership launch.

Hardware Performance Setup
Ryzen AI Max+ 395 (128GB unified) Fastest consumer path — runs unquantized in unified memory AMD ROCm + Ollama or vLLM with ROCm backend
Radeon RX 7900 XTX (24GB) Q5/Q6 quant fits comfortably, strong inference speed ROCm 6.x + Ollama or llama.cpp with ROCm
Radeon RX 7800 XT (16GB) Q4 quant, good for chat and coding Same ROCm stack, slightly lower throughput

AMD & Muse Glimmer — FAQ

Why did AMD partner with Meta on Muse Glimmer?
AMD wants to prove that local AI does not require NVIDIA hardware. Muse Glimmer running well on Radeon and Ryzen AI Max+ is a showcase for AMD's ROCm software stack and their vision of AI on consumer PCs.
Do I need ROCm to run on AMD?
For best performance, yes — ROCm 6.x enables GPU acceleration on Radeon cards. Ollama supports ROCm natively. On Ryzen AI Max+ with unified memory, the integrated GPU path also works.

See all hardware options or get the full setup guide.

Full Setup Guide Benchmark Scores