-
Beyond Cherry-Picking: Evaluating Text-to-Image Models
The HEIM Image Comparison Lab on Azure Container Apps provides a comprehensive framework for evaluating Text-to-Image models across fidelity, reasoning, safety, equity, and efficiency.
-
Microsoft Agent Framework: Hosted Agents in Microsoft Foundry with .NET
Microsoft Foundry offloads the infrastructure management for .NET agents, replacing manual Azure Functions hosting with a managed execution environment.
-
Copilot Studio | The Agent Your Flow Can't Call
Copilot Studio introduces two distinct harnesses-standard for rule-based agents callable via Power Automate and GitHub Copilot for reasoning-heavy workflows-shaping how agents are built, billed, and integrated.
-
Demystifying Dynamic AI Routing: Exploring the model router in Foundry Models with RouteLab
The model router in Foundry Models, paired with RouteLab, lets developers dynamically choose LLMs based on prompt complexity and performance needs, offering balanced, cost-effective, or high-quality routing modes with a local testing environment for visual evaluation.
-
Schema Validity Is Not Enough
Microsoft Foundry Local powers this robot task-planning prototype, which uses a deterministic governance layer to filter untrusted SLM proposals before execution.
-
When AI Stops Waiting for Instructions
Microsoft's Autopilots introduce persistent agents with identities, email, calendar, and organizational roles, automating tasks like release management within defined boundaries.
-
Claude Opus 5.5 comes to Microsoft Foundry for long-running coding and knowledge work
Claude Opus 5.5 in Microsoft Foundry improves focus, action communication, and token efficiency for complex coding and knowledge work in software development and enterprise applications.
-
GPT-6 Astra, Sol, and Luna: For production agents in Microsoft Foundry
GPT-6 Sol and Luna in Microsoft Foundry provide less noisy, more capable agents for efficient task completion, with Astra excelling in complex reasoning, Sol ideal for general purposes, and Luna optimized for high-volume data processing.
-
More Models, One Copilot
More Models, One Copilot showcases the launch of Anthropic’s Claude Opus 5.5 and OpenAI’s GPT-6 Sol in various Copilot apps, enhancing user options while Work IQ maintains relevance for each organization.
-
Microsoft Foundry Hosted Agents and MCP in Practice: Building Fibey Field Ops
Fibey Field Ops shows bridging model capabilities with operational tools using one hosted agent, five skill sets, eleven tools, and Azure Container Apps for a synthetic fiber-operations assistant built with Microsoft Foundry Hosted Agents and Model Context Protocol.
-
Transforming Solo Agents to Collaborative Teams: Choosing Multi-Agent Patterns in Microsoft Foundry
Choosing multi-agent patterns in Microsoft Foundry involves selecting orchestration strategies like sequential, concurrent, handoff, group chat, or custom workflows to manage agent collaboration efficiently for reliable, secure, and cost-effective AI systems.
-
Decoding Environmental Viromes: Amplifying Human Expertise with Microsoft Discovery
Microsoft Discovery speeds up public health insights from wastewater metagenomics by offering an easy-to-use app that combines sequencing data with scientific literature to lower computational and interpretation hurdles for researchers.
-
DeepSeek-V4.1-Flash is coming to Microsoft Foundry
DeepSeek-V4.1-Flash for Microsoft Foundry features a 20-layer causal encoder-decoder architecture with native image understanding, reduced KV cache usage to 890 bytes per token, and enhanced agentic task performance across coding, terminal, security, tool use, and automation benchmarks.
-
Engineering Agentic Recall Controls with MCP and Microsoft Foundry
Caldova Recall Control Tower shows how to separate AI reasoning from actions using MCP, Foundry, FastAPI, and Azure for validated pharmaceutical recall decisions.
-
Tokenomics: the four levers that decide your Azure OpenAI bill
Tokenomics details Azure OpenAI billing by emphasizing four key levers to reduce costs and improve efficiency in generative AI systems.
-
Insights in Foundry Turns Agent Traces into Action
Insights in Foundry automatically analyzes agent traces to uncover hidden patterns and failures, providing explanations and actionable recommendations without requiring custom evaluators.
-
Ship agents faster with expanded model choice, voice agents, and continuous optimization
Microsoft Foundry expands model selection, enables voice-enabled agents, integrates knowledge tools, optimizes performance with production data, and maintains existing investments.
-
Introducing voice agents in Microsoft Foundry
Microsoft Foundry introduces a unified platform integrating real-time speech, agent capabilities, deployment, observability, and evaluation for developing natural, responsive conversational AI across industries.
-
Microsoft Foundry Prompt Agents vs. Hosted Agents: What Are You Actually Choosing?
Choosing between Prompt Agents and Hosted Agents in Microsoft Foundry depends on whether you need direct control over execution-Hosted Agents let you run custom code for complex workflows, while Prompt Agents offer quick configuration with managed runtime.
-
From Camera to Canvas: How Multimodal GenAI Automates Shop Floor Instructions
The solution automates converting assembly videos into illustrated work instructions using Azure-based multimodal GenAI, reducing documentation effort and improving worker onboarding while preserving operational knowledge.
-
Your AI Gateway Is the Only Place Governance Actually Sticks
AI gateways centralize policy enforcement, preventing fragmented security and compliance issues as LLM deployments scale across an organization.