Google's Gemini 3 AI is here: How does it fare against ChatGPT, Grok and others

Google has officially launched its latest AI model, Gemini 3, marking a significant step forward in the realm of artificial intelligence. This new model, revealed on Tuesday, is touted as the pinnacle of AI technology, following the previous Gemini 2.5 Pro model that was considered the best for various applications. In recent benchmarks, Gemini 3 has not only outperformed its predecessor but has also established a commanding lead over competitors such as ChatGPT and Claude. According to Google, the Gemini 3 Pro model has achieved a remarkable score of 1501 in text-related tasks on the LMArena leaderboard, making it the top model and overtaking both Grok 4.1 Thinking and Grok 4.1. Additionally, Gemini 3 Pro has excelled in the WebDev category, surpassing GPT-5, which is indicative of its versatility and capabilities across various metrics. It now holds the number one spot in coding, mathematics, creative writing, and handling long queries on nearly all leaderboards. In an academic reasoning test known as Humanity's Last Exam, Gemini 3 Pro scored 37.5 percent, significantly outperforming GPT-5.1, which scored 26.5 percent, and Claude Sonnet 4.5 at 13.7 percent. Furthermore, in challenging math competition evaluations on MathArena Apex, Gemini 3 Pro achieved a score of 23.4 percent, while its rivals scored well below 2 percent. The model has also shown improvements in understanding computer screens, scoring 72.7 percent on the ScreenSpot Pro benchmark, again outpacing Claude Sonnet 4.5 and GPT-5.1, which scored only 36.2 percent and 3.5 percent, respectively. However, despite its impressive performance, Gemini 3 Pro did not dominate all coding-related tasks, falling short on the SWE-Bench Verified benchmark where Claude Sonnet 4.5 secured the top position with 77.2 percent, leaving Gemini 3 Pro in third place with 76.2 percent, just behind GPT-5.1. As the pace of AI model releases accelerates, the future might see Gemini 3 Pro challenged for its leading position sooner rather than later. While it currently stands out in numerous benchmarks, it's important to remember that these metrics may not fully capture the real-world performance of AI models, which ultimately hinges on user experience.