Fordel StudiosMicrosoft makes GPT and Claude collaborate in Copilot
What Happened
Microsoft added Critique and Council features to Copilot on March 30, 2026, enabling GPT and Claude to operate sequentially or in parallel within the same workflow. Multi-model configurations outperformed single-model baselines on internal benchmarks. The move makes multi-LLM orchestration a native Microsoft product feature rather than a developer pattern.
Our Take
Honestly, Microsoft just productized what we've been hacking together in LangGraph for the past year. Critique and Council — cute names, but what they're describing is a judge-executor loop and a panel vote. We've built both. The difference is Microsoft shipped it in something normals actually use.
Here's the thing: it working in a consumer product is different from it working in a Jupyter notebook. That's not a small gap.
The 'pick one model' era was already dying. Anyone serious was already mixing Claude for reasoning and GPT-4o for speed and cost. Now that's a first-class feature instead of a prompting trick you explain apologetically in your README.
What I'm actually watching: whether enterprise Copilot exposes this via API. If we can route to a Claude-reviewing-GPT pipeline programmatically, that's worth rebuilding around. If it stays locked in the UI, it's a checkbox for the deck.
Benchmark improvements are real but benchmarks always look good before production. The architecture is sound — the execution is the bet.
What To Do
Check the Copilot Studio API docs this week to see if Critique/Council is exposed programmatically before you build your own multi-judge pipeline on LangGraph.
Cited By
React