OpenAI is about to launch another big move!
Today, a brand-new image model, codenamed mona-lisa-1, officially launches on Arena.
Netizens who tested it firsthand were amazed, saying the image output of Mona-Lisa-1 is phenomenal and clearly surpasses GPT Image 2.
The screen appears more realistic, with a noticeable reduction in plastic-like appearance.
The new king of GPT Image is coming
Four months ago, GPT Image 2 made a stunning debut, instantly igniting the AI community.
At the time, Image 2 completely outperformed the Nano Banana Pro and has since held the throne as the king of AI image generation.
But now, a transformation has arrived.
An anonymous model codenamed "mona-lisa-1" has quietly entered the LM Arena blind leaderboard.
Netizen @AiBattle_ directly dug deep and ran its generated image through OpenAI’s verification tool—
Wow, they directly detected the SynthID watermark! Case solved—this mysterious superpower is another release from OpenAI.
It seems that this time, OpenAI is determined to go all in and outdo itself.
The plastic look is gone.
Developer Riccardo Wolf fed the same prompt to both models.
The output shows that Mona-Lisa-1 has better realism and reduced plasticity.
From left to right: Image 2 (medium); mona-lisa-1; Image 2 (high)
Take the generated live-action photo of Ultraman—the improvement in texture is particularly noticeable.
According to netizen Tim Jayas, the output of mona-lisa-1 has much richer detail.
In addition, when generating these highly evocative works of art, mona-lisa-1 demonstrates exceptional aesthetic quality.
Whether it’s the blending of colors, the layers of light and shadow, or the overall atmosphere, it has completely shed the rigid, algorithmic feel of past AI-generated art.
Infographic, more complex
GPT Image has always been highly skilled at generating complex infographics.
In practical tests, someone used Mona-Lisa-1 to directly generate a web UI with highly realistic visual effects.
Mona-Lisa-1 also handles the extremely detail-intensive "human organ dissection diagrams" with ease.
It can accurately generate everything from information-dense “GPT-5 one-page overview” to imaginative “exclusive self-portraits of ChatGPT.”
It can also mimic Einstein's handwriting and output the final result directly.
Some have claimed that ChatGPT's new image model has once again ended the competition.
A simple prompt can identify an anime wallpaper.
Plus, tons of amazing demos—just check out the images:
The "brain" hasn't been replaced.
Completed training last year
Careful developers have found that the training of Mona-Lisa-1 was likely completed last year.
@ivanainai conducted five date-related tests in a row, and the model's responses all remained stuck at 2025.
She then attempted to have the model process content involving the latest news, but all attempts failed. Of course, this was also partly because the anonymous version did not have internet access enabled.
Interestingly, a horizontal comparison reveals that GPT Image 2’s knowledge cutoff date is also December 2025.
In other words, based on the information revealed so far—
Mona-Lisa-1 is more like a new checkpoint on the same generation of foundation rather than a completely new brain.
In fact, some users who conducted multiple rounds of testing felt that Mona-Lisa-1 is merely the medium-sized version of Image 2.5.
On the other hand, the "-1" at the end of the code name almost explicitly indicates that new checkpoints will continue to emerge.
In other words, the version being tested now may not be the final form of the new GPT Image.
If Mona-Lisa-1 is truly just an intermediate checkpoint, then OpenAI's final official release could represent another major leap forward.
Only four months after GPT-4o reached the top, OpenAI has already begun breaking its own record.
At this rate, when the official version is finally released, the ceiling for AI-generated images will likely be raised even higher.
Next, we’ll just have to wait and see when OpenAI reveals the true identity of Mona Lisa.
Reference materials:
https://x.com/AiBattle_/status/2086405404368503240?s=20
This article is from the WeChat public account "New Intelligence Yuan," authored by ASI Revelation, edited by Peach.