OpenAI Steps Up with GPT 5.2, Overtakes Claude Opus 4.5 in GDPval-AA

 

There have been many criticisms about OpenAI and its models since GPT 5. But in face of competition from Gemini, the company stepped up to release GPT 5.2, which now leads across many benchmarks. It has even overtaken Claude Opus 5.2 in GDPval-AA, which measures performance on real-world, economically valuable tasks across multiple industries. There is a catch though: GPT-5.2 cost $620, compared to Claude Opus 4.5’s $608 and GPT-5.1’s $88.

GPT-5.2 just overtook Claude Opus 4.5 to achieve the highest score in GDPval-AA, a benchmark that focuses on performance in real-world economically valuable tasks

However, GPT-5.2 is also the most expensive model to run GDPval-AA: GPT-5.2 cost $620, compared to Claude Opus 4.5’s… pic.twitter.com/YsHTCKGp1K

— Artificial Analysis (@ArtificialAnlys) December 12, 2025

This is partially due to OpenAI raising prices but GPT 5.2 also uses more tokens. As Artificial Analysis explains, GPT 5.2 uses 25M tokens to run the 220 agentic tasks in GDPval-AA. You can find out more here.