Microsoft

OptiMind-SFT

OptiMind-SFT is a specialized 20-billion parameter model developed by Microsoft Research's Machine Learning and Optimization (MLO) Group. It automates the translation of complex decision-making problems, such as supply chain planning and resource allocation, into correct Mixed-Integer Linear Programming (MILP) formulations.

MAI-Image-2.5

MAI-Image-2.5 is Microsoft's latest AI image generation model, designed to create high-quality, photorealistic images from text or photo prompts. It offers precise control over edits, making it suitable for both casual and professional use. The model excels in text rendering, stylized illustration, and commercial imagery, and is integrated into Microsoft products like PowerPoint and OneDrive for Foundry users.

Copilot Cowork

Copilot Cowork is an advanced feature within Microsoft 365 Copilot, developed in collaboration with Anthropic, designed to handle long-running, multi-step tasks across Microsoft 365 applications. It enables users to delegate complex workflows to Copilot, which manages and executes them over time, providing continuous progress updates and allowing for user guidance or intervention as needed.

VibeVoice-ASR

VibeVoice-ASR is a unified speech-to-text model developed by Microsoft Research, designed to process up to 60 minutes of continuous audio in a single pass. It generates structured transcriptions that include speaker identification, timestamps, and content, with support for customized hotwords to enhance accuracy in domain-specific contexts.

Microsoft Agent Governance Toolkit

Microsoft released a critical set of 7 open-source packages on GitHub and PyPI designed to ensure autonomous agents follow security protocols. The toolkit includes a sub-millisecond policy engine and cryptographic agent identities, giving enterprises the tools to govern, audit, and secure AI agents operating in production environments.

Microsoft MAI Models

Microsoft AI Frontiers has launched three new world-class in-house AI models: MAI-Transcribe-1 (speech-to-text), MAI-Voice-1 (voice synthesis), and MAI-Image-2 (image generation). Built by Mustafa Suleyman's superintelligence team, all three models are now available via Microsoft Azure AI Foundry, offering world-class quality at competitive speeds and pricing independent of OpenAI.

Azure Skills

The Azure Skills Plugin is an integrated capability layer for AI coding assistants that bundles curated Azure expertise with MCP-backed execution. It enables agents to move beyond generic cloud advice by providing structured workflows, decision trees, and direct access to over 200 Azure service tools.

MAI-Voice-2

MAI-Voice-2 is Microsoft's advanced text-to-speech model that converts text into expressive, natural-sounding speech across 15 languages. It offers realistic expression, instant voice matching, and is designed for long-form content like audiobooks and podcasts.

MAI-Transcribe-1.5

MAI-Transcribe-1.5 is Microsoft's advanced speech-to-text model, delivering high-accuracy transcription across 43 languages, including English, French, Arabic, Japanese, and Chinese. It excels in challenging audio conditions, providing precise, domain-specific transcripts suitable for captions, call analysis, accessibility, and content workflows. The model supports entity biasing, enhancing recognition of industry-specific terminology, proper names, and other domain-specific terms, making it adaptable to various industries and applications.

dynwinrt

dynwinrt is a dynamic Windows Runtime (WinRT) API projection for Node.js and Electron that allows developers to call native Windows APIs directly from JavaScript or TypeScript. It eliminates the need for writing native C++ addons or complex build steps by reading .winmd metadata and invoking COM vtables at runtime via libffi.

MAI-Code-1-Flash

MAI-Code-1-Flash is a lightweight, agentic coding model developed by Microsoft to enhance developer productivity. Integrated into GitHub Copilot and Visual Studio Code, it assists in writing better code faster by providing efficient, real-time support for complex coding tasks.

winapp CLI

The Windows App Development CLI (winapp CLI) is a unified command-line interface designed to streamline the development of Windows applications. It simplifies complex tasks such as managing Windows SDKs, handling MSIX packaging, generating app identity, and configuring manifests and certificates for various app frameworks.

MAI-Thinking-1

MAI-Thinking-1 is Microsoft's flagship reasoning model, designed for complex problem-solving with competitive reasoning capabilities and top SWE-Bench Pro results at a mid-weight price. Built from scratch for serious math, coding, and real-world enterprise deployment, it stands among the strongest in its class.

run-assert-eval

run-assert-eval is an automated skill designed to streamline the responsible AI lifecycle for agents by integrating threat modeling, evaluation, and policy enforcement. It enables developers to discover risks, measure agent failure rates, generate runtime policies, and verify fixes through a unified, repeatable loop.

MAI-Cyber-1-Flash

MAI-Cyber-1-Flash is a specialized, compact AI model designed for cyber defense and vulnerability management. It is deeply integrated into MDASH, Microsoft's multi-agent vulnerability identification and remediation harness, to provide efficient, high-performance security protection.

ThinkingBox

ThinkingBox is an open-source sandbox and benchmarking framework designed to evaluate the reliability of AI agents in stateful business workflows. It separates the execution framework from the benchmark package, allowing for independent updates to the harness and test scenarios.

Agent Framework Go

Microsoft Agent Framework for Go is an open-source SDK that enables developers to build, orchestrate, and deploy production-grade AI agents and multi-agent workflows using the Go programming language. It provides a consistent foundation for agent development, offering support for various LLM providers, tool-calling, middleware, and observability features tailored for cloud-native applications.

Skill Recorder

Skill Recorder is a desktop application that captures your on-screen work sessions, including clicks, window switches, and optional narration. It leverages the GitHub Copilot CLI to reconstruct these actions into clear, intent-based steps, enabling the creation of reusable skills or automations for AI agents like Microsoft Scout and Copilot Studio.

tgrep

tgrep is a high-performance, trigram-indexed grep tool designed for rapid regex searching within massive codebases. By utilizing a client/server architecture with persistent indexing, it significantly reduces search latency compared to traditional tools like grep or ripgrep.

.NET Agent Skills

.NET Agent Skills is a curated collection of plugins and tools designed to provide AI coding agents with specialized domain expertise in .NET and C# development. The repository enables agents to perform complex tasks such as performance diagnostics, MSBuild optimization, Entity Framework data access, and framework migrations.

AutoSaddler

AutoSaddler is an open-source framework designed to automatically optimize LLM-agent harnesses by diagnosing execution traces. It improves agent performance by applying structured updates to prompts, tools, and middleware, ensuring that changes generalize across tasks rather than just fixing specific trajectories.

TypeScript 7.0

TypeScript 7.0 is a high-performance native port of the TypeScript compiler written in Go, designed to deliver up to 10x faster build and type-checking performance. By leveraging modern hardware through shared-memory multithreading and parallelization, it maintains strict compatibility with the original TypeScript logic while significantly reducing latency in development workflows.

MAI-Code-1.1-Flash

MAI-Code-1.1-Flash is a lightweight, agentic coding model designed to provide high-quality code generation with superior efficiency. It is specifically optimized for GitHub Copilot and VS Code, offering faster token streaming and reduced operational costs for engineering teams.

MAI-Transcribe-2

MAI-Transcribe-2 is a next-generation, high-accuracy speech-to-text model developed by Microsoft AI. It is designed to handle complex, real-world audio conditions across 60 languages, offering built-in features like diarization, timestamps, and domain-specific keyword biasing.