The open‑source llama.cpp library now prefetches model rows on Windows, improving inference speed on that platform.
NoteOossaPublished by Oossa: 1 min read
The llama.cpp project released version b11294 on 30 September 2026. The update adds row‑prefetching for Windows, meaning the library can load model data in advance to speed up generation. The change was contributed by Pranesh Gonegandla and merged without further modification. It also includes minor clean‑up and build fixes.
Windows users of llama.cpp will see faster model inference thanks to the new prefetch logic.
Last updated: ·Markdown
Models
Google’s new Guided Vision feature launches on Android phones, giving audio descriptions of what the camera sees, aimed at users with low vision.
Models
In a test, nearly half of participants thought the new AI video avatar was a real person, far above earlier systems.
Models
The companies will build GPT-Synopsys to operate Synopsys design tools. Early customer engagements are underway, but a general release date is not given.
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.
