Open-Source AI

Today's Stories

1. 1

   Google releases AI model that runs on regular laptops without internet connection

   Google DeepMind launched Gemma 4 12B on June 3rd, an AI model that can process text, images, and audio while running entirely on laptops with just 16GB of memory. The model is released under an open-source license (free for commercial use) and can work offline, making AI accessible to users who need privacy or lack reliable internet.

   Business travelers and security-conscious companies can now use advanced AI features on their laptops during flights or in secure environments without sending data to the cloud.

2. 2

   Microsoft unveils open-source framework for building AI agents at BUILD conference

   Microsoft announced its Agent Framework at BUILD 2026 on June 3rd, providing free development tools for creating AI agents (software that can perform tasks automatically). The framework includes Agent Harness for testing and Hosted Agents for cloud deployment, all released as open-source software.

   Businesses will be able to build custom AI assistants more easily and cheaply using Microsoft's free tools instead of hiring expensive development teams.

3. 3

   Nous Research launches desktop AI agent app for all operating systems

   Nous Research released Hermes Desktop on June 3rd, an open-source AI agent application that works across Windows, Mac, and Linux. The app allows users to run AI assistants directly on their computers under the MIT license (completely free to use and modify).

   Users can now install a powerful AI assistant on any computer without monthly subscription fees or cloud dependencies.

4. 4

   New interactive tool helps businesses choose the right hardware for AI models

   AgentSwarms published an interactive guide called 'Which GPU Runs Which LLM' that helps users determine exact hardware requirements for running AI models. The tool gamifies the process by letting users select model sizes and settings to instantly see what graphics cards they need.

   Companies planning AI deployments can now avoid costly hardware mistakes by knowing exactly what equipment they need before making purchases.

5. 5

   Microsoft releases open-source testing framework for AI applications

   Microsoft launched Adaptive Spec-driven Scoring for Evaluation and Regression Testing on June 2nd, an open-source framework that lets developers create AI behavior tests using simple text descriptions. The tool helps ensure AI applications work correctly before deployment.

   Software teams can now test their AI features more thoroughly and catch problems before users encounter them, leading to more reliable AI applications.