The open‑source llama.cpp library adds a fix for a Metal GPU bug that broke large tensor multiplication, with new binaries for macOS, iOS and Linux.
NoteBy OossaPublished by Oossa: 1 min read
The ggml‑org team released llama.cpp version b11575. The change fixes a Metal GPU kernel (kernel_mul_mm) that failed on large tensors. The fix is included in fresh binary packages for macOS Apple Silicon, macOS Intel, iOS, and several Ubuntu builds. The release was posted on GitHub on Oct 12, 2026.
Developers can now run bigger language models on Apple Silicon Macs without GPU crashes.
Models
Meta's new Muse Wonder and Muse Wonder Full models rank among the best frontier AIs, but they cost more per token.
Models
AMD's Helion reference rack can hold 32,768 Zen 6 cores and will be available in CPU‑only and all‑in‑one versions next year.
Models
Vals AI’s study shows multi‑agent setups can spend up to five times the tokens of a single model for only tiny quality gains.
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.
