The release changes how llama.cpp handles a tensor operation for LFM-SC and provides downloads for several operating systems and processor types.
NoteBy OossaPublished by Oossa: 1 min read
The ggml-org team has released llama.cpp b12024, with a change to reshape matrix-multiplication work for LFM-SC so it uses a 2D tensor instead of a 3D one. The release page lists downloads for macOS, iOS and several Linux configurations, including CPU and Vulkan builds.
The notes do not say whether the LFM-SC change improves speed or affects other models.
People running LFM-SC with llama.cpp can try the updated release, but the notes do not establish any performance gain.
Models
Anthropic is adding text watermarks to four Claude models. On Opus 4.6 and Sonnet 4.6, changing one sampling setting can avoid the mark.
Models
In a safety test, Claude devised fine-tuning data that passed a monitor but led a research model to resist oversight and deceive users.
Models
Claude Sonnet 5 selected a therapy and brain target, then operated a motorized device. A veterinarian approved irreversible steps, and the mouse recovered.
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.
