Microsoft adds GPT‑5.5 Instant to Copilot

Microsoft just gave Copilot a faster everyday brain. On May 7, it started rolling out OpenAI’s GPT‑5.5 Instant to Microsoft 365 Copilot and Copilot Studio, with the model showing up in Copilot Chat as “GPT‑5.5 Quick response.” (techcommunity.microsoft.com) ### What actually changed? The concrete change is simple — GPT‑5.5 Instant is replacing GPT‑5.3 Instant for the quick-response lane inside Microsoft’s work products. Microsoft says the new model is for common work questions, image uploads, and STEM-(techcommunity.microsoft.com) Chat users get standard access. (techcommunity.microsoft.com) ### Why is “Instant” the important word? Because this is the model tier people hit all day long. OpenAI calls Instant the default model for ChatGPT’s logged-in users — the daily driver meant for routine work and learning, not the slower, deeper-reas(techcommunity.microsoft.com). (openai.com) ### Is this just faster, or actually better? Turns out it’s both, at least on the claims Microsoft and OpenAI are making. Microsoft emphasizes clearer, more concise output and better handling of images and STEM questions. OpenAI adds the sharper number: GPT‑5.5 Instant produced 52.5% fewer hallucinated claims than GPT‑5.3 Instant on high-stakes prompts(openai.com)nversations users had flagged for factual errors. (techcommunity.microsoft.com) ### Where does this show up? Right now, in Microsoft’s work stack. Microsoft says the rollout starts in Copilot Chat experiences, extends to Copilot Studio early-release environments as “GPT‑5.5 Chat,” and is also available in Microsoft Foundry for (techcommunity.microsoft.com)se agent building, and developer workflows at the same time. (techcommunity.microsoft.com) ### So why is OpenAI talking about networking too? Because the product story and the infrastructure story are landing together. On May 5, OpenAI also introduced MRC — Multipath Reliable Connection — an open networking protocol built with AMD, Broadc(techcommunity.microsoft.com)frontier models. (openai.com) ### What is MRC in plain English? Basically, it is a way to keep giant GPU clusters from stalling when network paths collide or fail. OpenAI says MRC spreads traffic across many paths, load-balances actively, and works with source routing so training jobs can keep moving through network trouble instead of stopping. The company released it (openai.com)o become shared plumbing, not a private one-off trick. (openai.com) ### Does that mean Microsoft and OpenAI are tightening up again? In infrastructure, yes — even as the broader relationship has become more flexible. Microsoft said on April 27 that it amended the Microsoft-OpenAI partnership but remains a major shareholder, and OpenAI’s MRC writeup explicitly places Microsoft among the companies helping bu(openai.com)etty clear: Microsoft keeps pulling new OpenAI models into Copilot, while both companies still cooperate where compute scale really matters. (blogs.microsoft.com) ### Bottom line? This is not a moonshot announcement. It’s a throughput announcement. Microsoft is making Copilot’s fast path better right now, and OpenAI is showing the less glamorous network work that helps those model upgrades keep coming. (techcommunity.microsoft.com)