llama.cpp
llama-cpp-7e7dfac9·5 events·first seen 1mo agoAliases: llama.cpp
Co-occurring entities
More like this (12)
Recent events (5)
GGML and llama.cpp Join Hugging Face to Ensure Long-Term Progress of Local AI
GGML and llama.cpp, the foundational open-source libraries enabling efficient local inference of large language models, are joining Hugging Face. This move is intended to secure long-term development and sustainability of the projects that underpin much of the local/on-device AI ecosystem. The acquisition or integration represents a significant consolidation of key open-weights inference infrastructure under the Hugging Face umbrella.
New in llama.cpp: Model Management
llama.cpp has introduced new model management capabilities, as described in a Hugging Face blog post from the ggml-org. The post covers updates to how models are handled within the llama.cpp inference framework. This is a tooling update relevant to the open-source local inference ecosystem.
Introduction to ggml
This Hugging Face blog post introduces ggml, a C-based tensor library that underpins popular inference runtimes like llama.cpp and whisper.cpp. It explains ggml's design philosophy, quantization support, and how it enables efficient on-device inference for large language models. The post serves as an educational overview for developers looking to understand or build on the ggml ecosystem.
Simon Willison quotes Georgi Gerganov
Simon Willison shares a quote from Georgi Gerganov, the creator of llama.cpp. The body of the item is empty, so the specific content of the quote is unavailable. Georgi Gerganov is a significant figure in the open-weights inference ecosystem, making any substantive statement from him potentially relevant to tracking open-source LLM tooling trends.
FADA: Unified vision-language model for fetal ultrasound interpretation deployable on consumer smartphones
FADA is a unified vision-language model built on Qwen3.5-VL that performs clinical interpretation, classification, detection, and segmentation of fetal ultrasound images through a single pipeline without requiring external labels at inference. The system distills knowledge from four domain-specific foundation models using selective distillation, achieving 0.8820 mean Dice for segmentation and 0.7671 mAP@0.50 for detection, with expert validation confirming clinically acceptable outputs. Notably, the compressed 0.8B model runs entirely offline on a commodity smartphone (Qualcomm Snapdragon 7 Gen 1) in approximately 60 seconds, targeting diagnostic access gaps in low- and middle-income countries where trained sonographers are scarce. Code, models, and data are publicly released.