M365 Copilot Enterprise: Are the "GPT" labels native OpenAI models, and what is the true context window?

I am using M365 Copilot (Enterprise) and constantly hitting memory limits where the model loses the conversational thread during long document analysis.

Our tenant now displays a model selector with "GPT 5.5" under an explicit "OpenAI" label, including a "deeper thoughts" (reasoning) mode.

I have two technical questions pls:

- Model Authenticity: Are these actually the native OpenAI models running under the hood with identical reasoning capabilities, or is this a Microsoft-specific RAG pipeline just carrying the OpenAI branding?

- Context Window: Does this "GPT 5.5" mode offer a true context window comparable to native OpenAI/Anthropic models (e.g., 128k+ continuous recall)? The standard Copilot RAG drops data and forgets chat history far too quickly for heavy, multi-document synthesis.

I haven't found any technical facts on the architecture and the actual token limits behind these UI labels so I hope anyone can advise. Thanks!