Mysterious image generation model with OpenAI signals discovered in testing

A mysterious model designated mona-lisa-1 was discovered in the blind image testing list of the Language Model Arena. According to investigator reports, when images generated by this model were submitted to OpenAI's official content verification tool, origin signals related to OpenAI were detected. The model demonstrated superior performance to GPT-Image 2 in certain aspects of detail and realism, although some tests still identified occasional artifacts. To date, the model has only appeared in the Arena platform's blind testing section and has not been integrated into official channels such as ChatGPT or OpenAI's API. OpenAI has not released any official announcement, technical documentation, or confirmation of this model's relationship to existing products, leaving its exact identity unconfirmed.