OpenAI Mysterious image model raids Arena: the plastic feel disappears, suspected to be GPT Image 2

OpenAI mysterious image model raids Arena: the plastic feel disappears, suspected to be the GPT Image 2.5 medium cup version OpenAI A new image model codenamed mona-lisa-1 quietly appeared on the Arena (Arena), and was verified by netizens to have a SynthID watermark, confirming that it was created by OpenAI. Judging from actual test feedback, the new model's renderings are significantly more realistic, the lingering "plastic feel" in the past has been greatly reduced, and the details of the pictures are richer. It performs quite well in tasks such as artistic paintings and complex information graphics generation. What’s interesting is that the developers have uncovered a lot of clues. Training of the model may have been completed as early as last year, and its knowledge base deadline is consistent with the existing GPT Image 2, which is also December 2025. There are various indications that it is more like a new checkpoint growing on the foundation of the same generation. The outside world speculates that it is most likely a "medium cup" version of GPT Image 2.5, and more checkpoints may be launched in the future. If true, the official version is expected to further raise the ceiling of AI photography. via AI News (author: AI Base)