OpenAI has launched GPT-5.2, marking its third model release since August, following GPT-5 and GPT-5.1. This latest model includes a refined routing system and improved conversational capabilities, addressing previous critiques of cold responses. Although touted as a response to Gemini 3, OpenAI didn’t disclose direct performance benchmarks on its promotional site, focusing instead on GPT-5.2’s advancements over earlier versions and its performance in the new GDPval benchmark, which evaluates knowledge work across 44 professions. In recent tests, GPT-5.2 outperformed competitors, such as Gemini 3 Pro and Claude Opus 4.5, scoring 55.6% on SWE-Bench Pro and 92.4% on GPQA Diamond. It claims to complete tasks over 11 times faster than human professionals and with significantly reduced error rates. While independent benchmarks are pending, users can expect ongoing improvements in ChatGPT’s performance, particularly in coding tasks, as the landscape of AI continues to evolve.
Source link
