tech
April 1, 2026
Microsoft research tool uses Anthropic and OpenAI models
The multi-model method is designed to improve accuracy.
TL;DR
- Microsoft's AI research tool now uses models from both OpenAI and Anthropic.
- A new 'Critique' layer uses Anthropic's Claude to review OpenAI's responses for improved accuracy.
- The 'Council' feature allows users to compare responses from different AI models side-by-side.
- This multi-model approach aims to boost accuracy and reduce errors, with Microsoft's agent scoring 13.8% higher on a deep research benchmark.
- While beneficial, using multiple models can increase costs and slow down response times.
- Microsoft is also developing its own AI models, which may initially work in conjunction with external models.
Continue reading the original article