XNews

Open-source projects & model news

Open-source projects & model news

Multiple open-source releases and community tools were announced. - Mistral released Devstral Small 2: 24B params, local laptop runnable, competitive code performance. - Open Scouts: open-source automated web monitors that run AI scouts, summarize findings and push alerts; Docker-ready. - Meta news: report that Meta’s closed Avocado model will distill from Qwen, Gemma and GPT-OSS. - ScholArxiv: mobile arXiv reader with Gemini integration for paper summarization and chat. - AI/ML Bookshelf: open collection of 13 classic AI/ML books and learning resources.

Key insights

Mistral released Devstral Small 2 for local, high-performance usage

Mistral published two open-source models including Devstral Small 2, a 24B-parameter model designed to run locally on laptops while offering competitive code performance. The release emphasizes privacy and local development convenience for practitioners.

Open Scouts provides automated, customizable web monitoring with AI summaries

Open Scouts is an open-source system for deploying automated AI ‘scouts’ that monitor web sources, extract signals, generate concise summaries and deliver notifications. Built with Next.js and Supabase and deployable via Docker, it aims to help users continuously track topics without manual refreshes.

Reports say Meta's Avocado model will distill multiple open models

Reporting indicates Meta’s upcoming closed-source Avocado model will distill knowledge from a set of open and closed models (including Qwen, Gemma and GPT-OSS), reflecting Meta’s strategy to integrate diverse model advances into its proprietary offering.

Research and education tooling expanded with ScholArxiv and AI/ML Bookshelf

ScholArxiv provides a mobile-friendly arXiv reader with Gemini integration for summaries and chat; AI/ML Bookshelf aggregates 13 freely shareable AI/ML books as a learning resource for practitioners and students.

Source posts