Observed arrival · 2026-11-14
Sneo’s 70.6 GB local-model test
A Windows engine for running local mixture-of-experts models larger than GPU memory, using VRAM, system RAM, and CPU together.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Windows users running large local language models
- Worth noticing
- Sneo reports 93.3 tokens per second for a 70.6 GB model on an RTX 3090 with 24 GB VRAM and 32 GB RAM.
Field notes
Its benchmark table separates generation speed, prompt-reading speed, and time to first token across four prompt sizes; the page says each chart column is the median of four runs. The main test uses an RTX 3090, i9-12900K, 32 GB DDR5, and Windows 10, with a 64K context window and thinking enabled. The comparison alternated requests between Sneo and Strata, though Strata used its own app settings.
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
ƒJavaScriptBrowser-side code central
One card from the complete issue