OpenAI is testing the new generation image model, gpt-image-2, focusing on complex charts and realism

According to Pulse by Deepsight monitoring, OpenAI's upcoming next-generation image generation model has been tested among some ChatGPT users and has appeared under a codename on a third-party leaderboard, with some users referring to it as gpt-image-2. The model is expected to be officially released in the coming weeks.

Insiders mentioned two key improvements. First, the model's ability to handle complex graphics has significantly advanced. For example, accurately placing sticky notes or text on a whiteboard in a specific position, a task the old model consistently struggled with. Second, the outputs no longer exhibit an "AI feel," addressing past flaws such as overly perfect lighting and airbrushed faces.

Behind the new model is OpenAI's user growth pressure. ChatGPT's weekly active users recently plateaued at around 920 million, missing the year-end 2025 target of reaching 1 billion. Sam Altman issued an internal "code red" at the end of last year, triggered by the popularity of Google's image model Nano Banana, with improving image generation listed as one of the code red's top priorities.

OpenAI aims to recreate the early 2025 "Ghibli Moment," when hundreds of millions of users flooded ChatGPT to turn their and their friends' photos into Ghibli style. With enhanced text rendering capabilities, images can now be more reliably used in ads and educational charts. This is one direction where OpenAI is trying to close the gap with Anthropic, the latter recently surpassing OpenAI in revenue through its programming models.