Observed arrival · 2026-10-13
Measured Locally’s home-lab AI benchmarks
Alex publishes timed comparisons of local AI models running on hardware he owns, with results, test conditions, and related reporting.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People comparing hardware for running AI models at home
- Worth noticing
- The tests use the same prompts with model thinking switched off; the Spark GPU cap cuts peak power from 55.5W to 43.3W in one test.
Field notes
The benchmark table separates writing from reading (prefill) speed and dates the results. Its Spark figures use graphics chips capped at 2,200MHz; the author says an August test kept writing speed within 1 per cent while peak chip power fell from 55.5W to 43.3W. The homepage notes that uncapped Sparks would run somewhat faster, making the power-saving condition relevant when comparing the published numbers.
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
◉HumanPersonal, local, civic, or handmade
One card from the complete issue