Researchers have confirmed that GPT-6 is already in training, while OpenAI’s experimental model has achieved a remarkable milestone by competing in the 2025 International Math Olympiad (IMO). Performing at a gold medal level, the AI solved five out of six problems, earning 35 out of 42 points, under conditions identical to human competitors. This achievement showcases significant advancements in AI reasoning capabilities, capable of tackling complex problems that require extensive cognitive effort.
Evaluator Alexander Wei emphasized the model’s development from basic arithmetic to Olympiad-level challenges. Although the IMO-capable model represents a breakthrough, OpenAI clarified it is part of an internal research track and not slated for public release soon. Meanwhile, the anticipated GPT-5 will arrive shortly but will lack advanced math capabilities. Co-founder Yuchen Jin revealed that GPT-5 will feature a multi-model system for dynamic user interaction, highlighting the rapid evolution in AI technology leading to GPT-6.
Source link
