OpenAI Rumored to Be Testing New Image Model luna-lisa-alpha, Possibly Early Version of GPT Image 2.5

Multiple users have recently discovered an image model named luna-lisa-alpha on the LMArena testing platform, suspected to be OpenAI's next-generation image generation model under testing. Although not officially confirmed, external speculation suggests it could be an early version of GPT Image 2.5 or the next-generation GPT Image.

Based on test feedback, the new model shows notable improvements in character consistency, image realism, text generation capabilities, and noise control. Some testers comparing it with models such as GPT Image 2.0, Nano Banana Pro, and Grok Imagine believe that luna-lisa-alpha better maintains the same character's appearance, attire, and style across multi-round generations. However, some testers also noted that complex texture areas still exhibit repetitive texture issues. On the same day, OpenAI announced that the GPT Image 2 API preview supports transparent background generation, which can be used for product images, web design, and other scenarios. The exposure of multiple checkpoint names has also led the community to speculate that OpenAI may be conducting A/B testing. Whether this model will ultimately be officially released remains uncertain, pending official information from OpenAI.