The Oct 13 2026 release adds graph changes that avoid unnecessary LM‑head calculations when only pooled embeddings are needed.
NoteBy OossaPublished by Oossa: 1 min read
The llama.cpp library (maintained by ggml‑org) released version b11597 on Oct 13 2026. The update changes the computation graph so that when a context extracts pooled embeddings, the model no longer builds the large LM‑head logits layer. This reduces memory use and speeds up inference for embedding‑only tasks. The change also adjusts related graph nodes to keep the shape consistent.
Developers extracting embeddings can run llama.cpp models faster and on smaller hardware.
Models
Max routes requests to the most suitable model, cutting cost by up to 55% while improving performance in chat, code and work tasks.
Models
Google’s Gemini 4 Argon model is now in Devin Desktop and CLI at 75% off, and scores top on the FrontierCode 1.1 benchmark.
Models
The new Gemini 4 Argon AI model is now available globally via API and Google apps, priced at $2 per million input tokens.
Oossa · Newsletter
The week in AI, explained
Every Monday: the stories worth knowing, in plain language. Free, no spam.
