Microsoft has begun rolling out OpenAI’s GPT-5.2 model as a free upgrade for Copilot users on the web, Windows and mobile.
The model is offered through a new mode called “Smart Plus”, alongside the existing Smart mode based on GPT-5.1. The rollout, confirmed on 29 December 2025, puts OpenAI’s most advanced reasoning model in the hands of millions of users just weeks after its official launch on 11 December.
GPT-5.2 is a significant leap in AI capability. Its “Thinking” version matched or beat human experts in 70.9% of knowledge tasks across 44 occupations, compared with just 38.8% for GPT-5. It completes professional tasks such as presentations, spreadsheets and strategy documents more than 11 times faster and at less than 1% of the cost of expert human labour.
Improved performance on technical benchmarks
The new model also shows major gains on technical benchmarks. GPT-5.2 Thinking scored 55.6% on SWE-Bench Pro and 80% on SWE-Bench Verified, ahead of GPT-5.1 Thinking. It scored a perfect 100% on AIME 2025 maths problems and 92.4% on the GPQA Diamond science benchmark. GPT-5.2 also cut factual errors by about 38% compared with GPT-5.1, improving one of the persistent reliability challenges of large language models. It supports a 400,000-token context window with a knowledge cutoff of 31 August 2025, enabling analysis of very long documents and complex codebases.
Enterprise integration and access
Microsoft 365 Copilot subscribers have had access to GPT-5.2 since 10 December, and it is expected to reach all licensed users in the coming weeks. The integration connects GPT-5.2 to Microsoft’s Work IQ system, letting the model reason across workplace data such as meetings, emails and documents. Microsoft Copilot Studio has also given US customers building custom agents access to experimental versions of GPT-5.2.
OpenAI released GPT-5.2 on 11 December after CEO Sam Altman issued an internal “code red” to accelerate the launch amid competitive pressure from Google’s Gemini 3. The model comes in three versions: “Instant” for quick answers, “Thinking” for complex reasoning and “Pro” for maximum accuracy on difficult questions.
