Microsoft 365 Copilot’s Researcher agent goes multi-model

Microsoft 365 Copilot's Researcher agent will take a multi-model approach where Anthropic's Claude will review responses drafted by OpenAI's ChatGPT in a feature called Critique. Microsoft also added a feature called Council, which will show side-by-side comparisons of responses from different models.

Critique will be the default in Researcher when auto is selected as the model picker.

The moves illustrate three things:

1. Microsoft emphasizes diversity of models in its offerings with accuracy as the key benchmark.
2. The company is diversifying from OpenAI and adopting a multi-model approach in its services.
3. Microsoft's bet is that the differentiation is at the context and orchestration layer. Microsoft's Work IQ effort is another effort that highlights context as a key enterprise enabler.

Microsoft launches new E7 suite to integrate AI agents, Work IQ

Separately, Microsoft said Copilot Cowork, powered by Anthropic's Claude is available in the Frontier program.

In a blog post, Microsoft said:

"Critique is a new multi model deep research system designed for complex research tasks. It separates generation from evaluation and utilizes a combination of models from Frontier labs including Anthropic and OpenAI. One model leads the generation phase, planning the task, iterating through retrieval, and producing an initial draft, while a second model focuses on review and refinement, acting as an expert reviewer before the final report is produced. Our evaluations show that this architecture exceeds traditional single model approaches and delivers best in class deep research quality."

According to Microsoft, one model will act as the reviewer and gauge source reliability, completeness and evidence. The other model will do the report.

As for Council, Microsoft said it's an alternative approach to Critique. Council will run Anthropic and OpenAI at the same time with each model producing a report. Once reports are generated, a judge model evaluates the reports.

Although Microsoft's research agent tools are notable, the approach is likely to be adopted by other hyperscalers such as Amazon and Google Cloud that offer multiple models.