Microsoft Made GPT and Claude Work Together—And the Result Beats Every AI Research Tool Out There
Major tech companies have released advanced AI research agents, each claiming superiority. Microsoft has introduced two new features for its Copilot Researcher tool—Critique and Council—that address limitations in single-model research by combining multiple AI models. Critique uses OpenAI’s GPT to generate an initial report and Anthropic’s Claude to review it for accuracy and quality, resulting in higher performance, with Copilot scoring 57.4 on the DRACO industry benchmark, outperforming all competitors. Council, meanwhile, has GPT and Claude independently produce reports that are compared by a third model, which summarizes agreements and differences. Critique is the default, while Council is an optional mode. These features are available to Microsoft 365 Copilot users in the Frontier early-access program. Microsoft’s strategy is to integrate multiple models, believing orchestration among top systems delivers superior research results compared to relying on any individual AI model.
