According to a blog post published by tech media outlet BleepingComputer on December 29, Microsoft has begun rolling out the GPT-5.2 model to Copilot users across web, Windows 11, and mobile platforms, branding it as the new "Smart Plus" mode.

This update is offered as a free enhancement, launching with the fresh "Smart Plus" option that coexists alongside the previously released "Smart" mode based on GPT-5.1.

The version integrated into Copilot is officially designated as the "Thinking" variant of GPT-5.2, with Microsoft highlighting its significantly improved capability in handling complex tasks.

Equipped with this model, users can now accelerate practical workflows such as building spreadsheets, creating presentations, writing and reviewing code, interpreting lengthy documents, and processing images—officially positioning it as an expert-level assistant for specific office-related duties.

In evaluations of knowledge-based work tasks across 44 professions (GDPval), GPT-5.2 Thinking matched or surpassed human domain experts in 70.9% of test cases—an improvement from GPT-5's prior performance at 38.8%.

Benchmark results from leading industry standards further demonstrate GPT-5.2’s dominant technical edge. On SWE-Bench Pro, a software engineering assessment, the model achieved a score of 55.6%, while reaching 80% on SWE-bench Verified—both figures surpassing those of GPT-5.1 Thinking.

GPT-5.2 also delivered outstanding results in academic and reasoning benchmarks: achieving a perfect 100% on AIME 2025 (American Invitational Mathematics Examination), scoring 92.4% on GPQA Diamond, 88.7% on CharXiv Reasoning (including Python), and delivering substantial gains in the ARC-AGI challenge.