Follow the latest coverage, related explainers and connected technology stories.
A joint demonstration by JetBrains and the Rust Foundation explains how to use the Rig library to build AI agents in Rust, with support for model providers, tools, RAG, and local models. The material also highlights an approach to testing model integrations by recording HTTP requests and replaying them in a continuous integration environment.
Alibaba’s open Qwen3.8-27B model offers multimodal capabilities, coding, and software-agent functions in a size that can run locally after being compressed to about 17 gigabytes. Despite strong performance indicators, its high token consumption and slow inference impose a practical trade-off between quality and speed.
bartowski presents an extensive experiment to improve the calibration data used with imatrix when quantizing models through llama.cpp, concluding that the greatest gains appear at low bit rates, particularly in mixture-of-experts (MoE) models. The new dataset is smaller and more focused, but combines diverse texts with tool-use conversations formatted for tool calling.
Liquid AI announced the availability of the LFM2.5-2.6B and LFM2.5-2.6B-Base models on Hugging Face, with support for tool calling and multi-step workflows locally. The model targets running fast agents on computers and phones, while using less than 2.5 gigabytes of memory.