Observed arrival · 2026-11-19
Loom moves model ports outside the inference engine
An experimental compiler, runtime, and serving stack for fetching, optimizing, and running community models without modifying the inference engine.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Inference developers testing AMD hardware
- Worth noticing
- The model ranking uses a geometric mean across decode, prefill, and four-stream workloads, with results from Strix Halo tests.
Field notes
The homepage lays out a three-command path: install hrx-loom, fetch a model package, and launch it with loom-serve. Its charts compare Loom and llama.cpp across decode, prefill, four-stream workloads, and context lengths; the page says the ports were agent-tuned and vary in optimization so far. The stated hardware scope for the model results is Strix Halo.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue