OpenAI appears to be quietly testing what it internally refers to as Image V2, a next-generation image generation model that surfaced on LM Arena in three distinct variants.
- packingtape-alpha
- maskingtape-alpha
- gaffertape-alpha
UPD: Images were pulled back from the Arena during the weekend but still can be encountered on ChatGPT.
Some ChatGPT users have reported gaining permanent access to the new model, while others are seeing its outputs through an A/B testing framework where they are asked to choose between competing results. This follows a familiar playbook for OpenAI, the company used the same Arena-based blind testing approach in December 2025 when it previewed models codenamed Chestnut and Hazelnut, which ultimately shipped as GPT Image 1.5 just weeks later.
Early impressions suggest Image V2 represents a meaningful step forward. Testers highlight its ability to render realistic UI interfaces with correctly spelled button text, a longstanding weakness in AI image generators, along with strong prompt adherence and compositional understanding. Comparisons to Google’s Nano Banana Pro are already circulating, with some users finding Image V2 competitive in areas where OpenAI’s current model still trails Google’s offering, which has held the top spot on the LM Arena leaderboard for months. OpenAI has been operating under what CEO Sam Altman described as a “code red” posture since Google’s Gemini 3 and Nano Banana Pro began eating into its market position in late 2025, and a strong Image 2 release would be a direct answer to that pressure.
Some Image V2 generations from X community: Elania, Can, levelsio, Angel, Flowers
The key question is whether OpenAI will maintain the model’s current quality at launch or dial it back for cost and safety reasons, a pattern the company has followed before. Pricing will also matter considerably, given that GPT Image 1.5 already undercuts its predecessor by 20% on API costs. There has been no official announcement, and the A/B testing phase could last from a few days to several weeks, depending on prior release cycles. Designers, marketers, and developers who rely on ChatGPT’s image capabilities stand to benefit most, particularly those working on UI mockups and commercial layouts where text accuracy is critical.
