llama.cpp Release b11364 Adds Support for Nimble Decision Model
Open-source LLM inference engine llama.cpp has released version b11364, adding support for the Nimble decision model via pull request #29844. The update also delivers updated pre-compiled binaries across multiple operating systems and backends, including CUDA 12/13, ROCm, Vulkan, and Snapdragon. Adding support for specialized models like Nimble enables users to efficiently execute fast, structured text classification and typed decision-making directly on local hardware. This release continues llama.cpp's trend of expanding compatibility with diverse neural network architectures beyond standard chat models. Nimble processes text against structured schemas to output typed decisions along with confidence scores for each answer token. The release notes also indicate that Arm KleidiAI acceleration support for macOS Apple Silicon builds remains disabled.
## BACKGROUND
llama.cpp is a high-performance C/C++ library designed for local LLM inference across diverse CPU and GPU architectures. Nimble, developed by Bespoke Labs, is a specialized decision model that scores text directly against user-defined schema choices rather than generating long-form free text.