Microsoft Copilot Researcher Adds Multi-Model Architecture Using GPT and Claude

Microsoft has expanded its Copilot Researcher tool with dual new features that combine OpenAI's GPT and Anthropic's Claude simultaneously for research tasks, according to a blog post announcing the Copilot Cowork platform.

Critique Feature: Sequential Model Refinement

The Critique feature generates initial responses using GPT, then refines them through Claude. Microsoft claims this architecture "creates a powerful feedback loop that delivers higher-quality results across factual accuracy, analytical breadth, and presentation."

The company states that Researcher's process mirrors "academic and professional research settings." According to Microsoft, the upgrade scores higher than Perplexity's Deep Research models on the Deep Research Accuracy, Completeness, and Objectivity benchmark—though specific benchmark numbers were not disclosed.

Model Council: Comparative Analysis

Alternatively, users can select Model Council to receive side-by-side responses from both Anthropic and OpenAI models. The feature includes a report highlighting where the two models agree and disagree, allowing researchers to evaluate multiple perspectives on complex queries.

Microsoft designed Researcher specifically for multi-step research tasks, distinguishing it from standard Copilot. Anthropic independently operates a Research feature using multiple Claude agents for similar purposes.

Availability and Context

Both Critique and Model Council features are currently available exclusively in Microsoft 365 Copilot's Frontier program, which functions as an early access testing ground for the company's AI innovations. No general availability date has been announced.

The move reflects a broader industry trend of combining complementary AI models to offset individual model weaknesses. Neither OpenAI nor Anthropic has publicly objected to this architecture, suggesting potential partnership or licensing arrangements, though details remain undisclosed.

What this means

Microsoft is positioning multi-model research workflows as enterprise standard practice rather than single-model reliance. This approach benefits organizations requiring high-confidence outputs on complex research tasks. The lack of published benchmark numbers limits independent verification of claims versus Perplexity's comparable offering. Availability remains gated to early access users, suggesting Microsoft is gathering feedback before broader rollout.