On Wednesday, Ethan Mollick, a Wharton associate professor and AI researcher, revisited his Otter Test to evaluate AI image generators. Using the adorable otter as a benchmark, Mollick tests the AI models’ ability to accurately generate images based on otter-related prompts. Testing the ChatGPT Images 2.0, now accessible to all ChatGPT users, Mollick expressed surprise at the exceptional quality of the results, indicating that Images 2.0 successfully passed his Otter Test.
Mollick’s positive experience with the latest AI image generators seems to be a shared sentiment. Following OpenAI’s launch of the new tool, touted as a cutting-edge image model capable of handling complex visual tasks, it has garnered widespread acclaim. Mollick, an early beta tester of the tool, attested to its impressive performance, highlighting its ability to generate diverse content like text, slides, and academic papers effectively.
The buzz surrounding the new ChatGPT image generators stems from their unique intelligence. Unlike conventional tools, these generators possess cognitive capabilities, enabling them to “think” and utilize general intelligence. OpenAI emphasizes that this general intelligence, based on data up to December 2025, significantly enhances Images 2.0’s proficiency in following detailed instructions, rendering complex text, accurately placing objects, supporting various aspect ratios, and generating multilingual images.
The exceptional abilities of the ChatGPT image generators have garnered widespread admiration for their advanced intelligence and output quality. This leap in AI technology reflects in the rankings on Arena AI, a platform where AI image generators compete. GPT Image 2 currently holds the top position by a considerable margin, dominating the Text-to-Image category with a record-breaking lead.
While words can describe an image tool’s capabilities, the real impact lies in visual representations. Notable creations by ChatGPT include an image of a mound of rice with the word “wow” inscribed on a grain, showcasing the tool’s creative potential.
Another striking image generated by ChatGPT portrays two individuals in a style reminiscent of the 80s and 90s, exhibiting remarkable realism and attention to detail.
Despite its impressive advancements, ChatGPT Image 2.0 is not flawless. Mollick acknowledges certain limitations, noting that while the images are of high quality, editing can be challenging, requiring multiple attempts to achieve desired changes.
Overall, the enhanced AI image generation capabilities of ChatGPT have drawn praise for their improved quality and usability compared to previous models. Users commend its ability to accurately depict and relate objects, render dense text, and handle textual elements within images effectively.
– Ends