OpenAI has once again raised the bar for artificial intelligence in highly technical domains. The company’s recently released GPT-5.2 Pro model has secured the top position on the rigorous FrontierMath leaderboard, demonstrating unprecedented capability in solving complex mathematical problems.
Released on December 11 as part of the new GPT-5.2 series, the Pro version achieved a stunning 29.2% accuracy on FrontierMath Tier 4. This benchmark tier is renowned for its extreme difficulty, consisting of unpublished challenges in advanced fields such as analysis and algebraic geometry problems designed to stump even human specialists for days.
A Massive Leap in Six Months
The significance of this score is highlighted not just by its current standing, but by the rapidity of the progress it represents. The 29.2% figure marks a massive jump from just six months ago, when similar models scored only 4% on these types of tasks.
Also Read: Fal.ai Open-Sources Ultra-Fast FLUX.2 Turbo Image Generator
Furthermore, this achievement places OpenAI significantly ahead of its primary competitors in this arena. The GPT-5.2 Pro score beats Google DeepMind’s Gemini 3 Pro Preview by over 10 percentage points.
Reactions from Leaders and Academics
The breakthrough has elicited strong reactions from both OpenAI leadership and early users in academia. OpenAI President Greg Brockman interprets this result as proof that the company is nearing the creation of AI that can genuinely excel in specialized technical fields.
Users testing the model for rigorous academic work have corroborated this assessment. Physicists have praised the model’s potential for aiding research, while Acer, a math student at Cambridge, described GPT-5.2 Pro as the strongest AI available for proof-writing.
Also Read: Claude Is Now Inside Your Browser | Here’s What Changes for Everyday AI Users
