What it is
Desktop chat app that also serves an OpenAI-compatible API (documented default http://127.0.0.1:1234/v1). Runtimes are bundled llama.cpp engines the app downloads (CPU, Vulkan, CUDA, ROCm).
Architecture in plain English
GUI + local server. You pick a runtime in Settings. It is not this site’s Vulkan llama.cpp tree. Pointing other harnesses at :1234 is the documented OpenAI path; pointing LM Studio at our :8080 is a current-UI check.
What it does well
A GUI on Windows or a first desktop when you refuse to compile — if a GPU meter moves.
What it does badly
Reproducing this site’s HIP/ROCm notes; silent CPU.
ROCm / Radeon
AMD’s LM Studio playbook says the app offers Vulkan and AMD ROCm runtimes. Older 0.2.31 “extension pack” scripts are version-specific — do not run them on a current build. gfx1201 completeness of the bundled ROCm engine is not documented as our daily driver.
Vulkan. Yes — the honest cross-vendor path. Prefer it over a mystery ROCm blob until the meter moves.
Supported quants
GGUF in the app library. Floor is Q4.
House floor is still Q4. Prefer Q5. See quants and thecalculator.
Install on an R9700
# Linux AppImage from https://lmstudio.ai/download?os=linux
# Settings → Runtime → download Vulkan (or ROCm) llama.cpp engine.
# Do not take a CUDA runtime on a Radeon.
Run
# Documented server: http://127.0.0.1:1234/v1
curl -s http://127.0.0.1:1234/v1/models
Tuning flags
If it is slow, assume CPU until a meter moves. Prefer the llama.cpp you built and point harnesses at :8080.
When to choose it
GUI convenience, not the house stack.
Windows-only, or a desktop, and you accept “no data” for tok/s.