The competition between AI systems ChatGPT and Gemini has intensified, with recent benchmark results indicating that ChatGPT, developed by OpenAI, leads in critical areas of performance. While both systems are highly capable, recent evaluations reveal significant advantages for ChatGPT, particularly in reasoning, problem-solving, and abstract thinking tasks.
Understanding the nuances of AI performance is essential, as rapid advancements can shift the landscape overnight. For example, in December 2025, speculation circulated regarding OpenAI’s position in the AI race, but the release of ChatGPT-5.2 soon after showcased its resurgence in capabilities. The evaluation of AI systems has become more complex, given the diminishing differences between prominent models like ChatGPT and Gemini, developed by Google.
ChatGPT Excels in Rigorous Testing
One of the primary benchmarks where ChatGPT has demonstrated its superiority is the GPQA Diamond, which assesses PhD-level reasoning in scientific disciplines. This benchmark employs complex, Google-proof questions that require deep reasoning rather than simple recall. ChatGPT scored 92.4%, slightly ahead of Gemini 3 Pro, which achieved 91.9%. To provide context, a typical PhD graduate is expected to score around 65%, while non-expert humans average just 34%.
Another critical evaluation is the SWE-Bench Pro (Private Dataset), which measures an AI’s ability to solve real-world software engineering problems sourced from GitHub. In this challenging benchmark, ChatGPT successfully resolved approximately 24% of issues, while Gemini managed to resolve only 18%. These results highlight the ongoing challenges AI systems face in matching human expertise, as humans typically solve all tasks successfully in this benchmarking context.
Abstract Reasoning Capabilities Compared
In the realm of abstract reasoning, the ARC-AGI-2 benchmark serves to evaluate AI’s performance in identifying patterns and applying learned concepts to new examples. Here, ChatGPT-5.2 Pro achieved a score of 54.2%, significantly outperforming Gemini 3 Pro, which scored only 31.1%. This underscores ChatGPT’s strength in handling tasks that require intuitive understanding and general fluid intelligence.
Despite these strong results for ChatGPT, it is essential to recognize that benchmarks can change rapidly, and performance can vary based on updates to the AI models. The current analysis focused on the most recent versions—ChatGPT-5.2 and Gemini 3—specifically their Pro editions, which generally rank higher in evaluations.
While ChatGPT currently leads in specific benchmarks, Gemini also has its strengths in various tests where it outperforms ChatGPT. For instance, Gemini has fared better in assessments such as SWE-Bench Bash Only and Humanity’s Last Exam. This article primarily highlights three benchmarks that showcase ChatGPT’s strengths to provide a balanced perspective on AI capabilities.
AI benchmarking remains a complex field, with multiple methodologies available for comparisons. While subjective studies like LLMArena aggregate user preferences effectively, this report aims to present a more objective analysis based on quantifiable data.
As these AI systems continue to evolve, ongoing evaluations will be necessary to determine their standing in the rapidly changing landscape of artificial intelligence. Current trends indicate that ChatGPT is leading in specific areas, but the competition remains fierce, and future updates may alter these dynamics.
