llama.cpp Releases Version b10486 with LFM2 Image Tiling Fix
llama.cpp has released version b10486, which introduces a bug fix for the LFM2 image tiling threshold. The update resolves issues related to image tiling and includes refactored testing, with specific fixes applied for Windows environments. This release ensures correct processing of visual inputs for Liquid AI's LFM2 models when running locally via llama.cpp. Maintaining compatibility and fixing bugs for edge-optimized models is crucial for developers deploying multimodal AI on consumer hardware. The fix was implemented in pull request #27057 and co-authored by a Hugging Face contributor. The release also provides pre-built binaries for various platforms, though certain builds like macOS with KleidiAI and Ubuntu with ROCm are currently disabled.
## BACKGROUND
llama.cpp is a popular open-source project that enables efficient LLM inference on local hardware using C/C++. LFM2 (Liquid Foundation Models 2) is a family of hybrid, lightweight models developed by Liquid AI designed specifically for edge AI and on-device deployment.