I asked five models
to make art.

I started with Astra Pro, which generated the first six images. The results established the visual direction for the collection and gave the other models something substantial to be measured against.

Astra's work felt confident and immediately usable. The images helped shape the look of the redesigned site, and they became an accidental reference point for what followed.

Claude Pro produced the next six images, taking a noticeably different route. Rather than treating the task as a conventional image-generation prompt, Claude created the artwork with code.

That led to a collection of modern abstract pieces. They were not all equally successful — some were stronger than others — but several had a distinct visual personality. The work felt more designed than illustrated: geometric, experimental, and deliberately synthetic.

Claude's contribution was a useful reminder that “AI art” does not have to mean one particular generation method. Code can be a creative medium too, especially when the goal is abstraction, patterns, composition, or motion-inspired design.

Grok Free generated six images of its own. I had specifically asked Grok to look at AiArtNow.com, so it made sense for the site to influence the result.

What amused me was how strongly Grok seemed to be inspired by Astra Pro's images. The resemblance was more noticeable than I expected. Grok had looked at the website as requested — but in doing so, it appeared to absorb some of the visual language Astra had already established.

That made the comparison less like a controlled contest and more like a conversation between models. One model set a direction, another interpreted it, and the influence became part of the story.

Google had the most difficult time with the assignment. I had to explain the purpose of the site, what the redesign was trying to accomplish, and why the images mattered. The process required more back-and-forth than I expected.

Still, I wanted the comparison to be honest, so I shared the text and images Google produced rather than quietly leaving the weaker results out. A fair comparison should include the awkward attempts as well as the impressive ones.

The roughness was informative. It showed that creative performance is not just about whether a model can produce an image. It is also about whether the model understands the brief, connects the visual work to the surrounding project, and can adapt when the first result misses the mark.

Meta AI went last, and I went in with modest expectations — it's the free tier, not a paid or pro model. Six original concepts, generated as an originality test rather than a copy of anything else on the site.

I was wrong to expect less. I'm genuinely impressed by the quality, and I'd put these images among the best in the whole collection — a loom weaving starlight into a nebula tapestry, an acorn cracked open to reveal a glowing city inside cupped hands, a canyon where spoken words become glass birds. The compositions felt considered rather than generic, and nothing about them read as a free-tier afterthought.

It's a useful reminder that price tier and output quality don't always move together. The most expensive model in this comparison wasn't the one that impressed me most.

The models did not behave like interchangeable art tools. Each one brought a different creative process:

1. Astra Pro

   Delivered a strong visual foundation.

2. Claude Pro

   Used code to make modern abstract art.

3. Grok Free

   Responded to the site and visibly echoed Astra's direction.

4. Google (Gemini)

   Needed more explanation and produced a rougher set of results.

5. Meta AI (Free)

   Went last, on the free tier, and turned in some of the strongest images of the set.

There was no single winner in every category. The most polished images were not necessarily the most interesting, and the least successful attempts still revealed something about how a model interprets creative instructions.

That is the point of this collection. AiArtNow.com is not only a gallery of attractive images. It is also a record of how different systems think about art, context, imitation, abstraction, and the limits of a brief.

The images are the evidence.
The differences between them are the experiment.