llama.cpp adds faster CPU GELU kernels in v b11786

The open‑source LLM runtime updates its CPU math to compute GELU functions in full precision without a lookup table.

NoteBy OossaPublished by Oossa: October 21, 2026 at 8:38 AM UTC1 min read

The ggml‑org/llama.cpp project released version b11786 on 2026-10-21. The update adds new CPU kernels that compute the f32 gelu_quick and geglu_quick functions directly, removing the need for an fp16 lookup table. Tests for these kernels were also added, and the old inf/nan test cases were dropped. Pre‑built binaries for macOS, iOS, and several Linux flavours are now available for download.

Why it matters

Developers can run LLMs faster on CPUs because the new kernels avoid the slower lookup table.

Sources & references

1 sources

llama.cppSource published: Oct 21, 2026Read original source

Published by: OLY Europe LTDReport an error

Markdown

Was this article useful?

llama.cpp adds faster CPU GELU kernels in v b11786

Why it matters

Sources & references

Read next

Canva launches its own image generation model, Canva Craft

AWS upgrades Amazon Polly’s generative speech engine

Uber expands ChatGPT Work to all 30,000 employees