Microsoft 365 Copilot Introduces Critique and Council Modes

Microsoft 365 Copilot Researcher has introduced multi-model-based Critique and Council modes.

Critique mode is a dual-model system in which two models collaborate. The first model performs research drafting, planning, and information summarization, while the second model reviews the credibility of sources and the accuracy of evidence and supplements the content. According to evaluation results on the DRACO benchmark, this approach improved overall research quality by 13.88% compared to Perplexity's Claude Opus 4.6.

Council mode simultaneously utilizes models from Anthropic and OpenAI to generate reports. A Judge model then compares and analyzes the outputs of both models, summarizing their commonalities, differences, and each model's unique insights.

These features are currently being rolled out first to professional and enterprise users through the Microsoft 365 Copilot Frontier program. Microsoft aims to establish a new standard for enterprise research assistance tools by integrating its own models with third-party models.