OpenAI is Reportedly Testing Its Next-Gen Image V2 Model in ChatGPT & Arena

OpenAI is quietly testing a brand new visual generation model across different platforms. Users recently spotted the software, currently called Image V2, appearing on the popular Arena website under some unusual codenames. While the company pulled the model from the public leaderboard over the weekend, a select group of people can still access it directly inside ChatGPT through an ongoing testing phase.

Here’s how the trial testing phase works

The artificial intelligence firm is using a familiar strategy to test the waters. It asks regular users to generate pictures and then choose the best result without knowing which software actually made it. This blind testing method gives the company real feedback on what people actually like.

It did the same thing late last year before officially launching its current standard image software just a few weeks later. Right now, some lucky accounts have permanent access to this upcoming version, while others only see the new results mixed into regular prompts.

OpenAI is fixing old text spelling problems

Early reactions from testers show that Image V2 is a serious upgrade. One of the biggest complaints about generated art is how it handles written text. The old versions often spit out garbled letters or nonsense words. The new software fixes this annoying issue. People say it can generate realistic user interfaces and buttons with correctly spelled words. It also follows complex instructions much better than before.

The tech brand needs a big win to stay competitive. Google currently dominates the top spot on the public leaderboards with its own visual tools. A strong release here could help the creator of ChatGPT reclaim its crown. The main question left is whether the public release will look as good as these test runs.

The firm sometimes lowers the quality of its products to keep operating costs down or improve safety. There is no official release date yet, but OpenAI is already excited about the potential for building accurate UI mockups and layouts.