Not ChatGPT, This is the AI Model with the Highest IQ

Companies in the artificial intelligence (AI) sector, such as OpenAI and Anthropic, continue to compete in launching increasingly intelligent AI models. Each time they release a new model, they often publish benchmark results to prove the superiority of their latest product. A similar practice is followed by Tracking AI, a platform dedicated to testing and comparing the performance of various AI models. This platform measures the intelligence level of AI models using IQ tests, with the results then visualised by Visual Capitalist, an independent media company known for its data infographic presentations. The testing was conducted in April 2026 using the Mensa Norway IQ test, involving at least 26 AI models from various companies in several versions. The results show extremely tight competition. Grok-4.20 Expert Mode and OpenAI GPT-5.4 Pro (Vision) both achieved the highest score of 145, so Visual Capitalist placed them jointly in first position. Gemini 3.1 Pro Preview followed behind with a score of 141. This achievement is an improvement compared to the 2025 testing, where the highest score only reached 135 points. Chinese AI models are no less competitive. Alibaba’s Qwen 3.5 became the highest-ranking Chinese AI model, placed in 10th position with a score of 130. Models such as DeepSeek R1 and DeepSeek V3 also made the list. Here are the full details. Non-vision models received questions converted into verbal descriptions, while vision models worked on the original image questions without modification. Tracking AI also applied special rules: if a model refused to answer, the same question would be repeated up to a maximum of 10 times, and the last answer would be used for evaluation. Nevertheless, these scores are only comparative and not an absolute measure of overall AI intelligence, as IQ tests only measure one aspect, namely pattern reasoning ability, as compiled by KompasTekno from Visual Capitalist.