Earlier than ChatGPT, there was DALL-E — the primary iteration of OpenAI’s now-famous text-to-image AI mannequin. It was launched in January 2021, giving the world an early style of the way forward for AI-generated imagery that was quickly to come back.
Right now, there is a lengthy roster of image-generating AI instruments to select from. Firms like OpenAI, Google, Midjourney, Stability AI, Ideogram, and Black Forest Labs all supply their very own fashions, every with its personal specific strengths and weaknesses.
(Disclosure: Ziff Davis, ZDNET’s mum or dad firm, filed an April 2025 lawsuit in opposition to OpenAI, alleging it infringed Ziff Davis copyrights in coaching and working its AI methods.)
Now, a brand new web site lets you preserve monitor of — and vote on — the leaders of this consistently evolving (and more and more crowded) class of AI fashions.
The way it works
Launched in beta final month, LMArena.ai ranks AI fashions “primarily based on their capacity to generate photographs primarily based on textual content descriptions,” in line with its homepage. The web site was initially based as a analysis initiative by way of UC Berkeley with the title Chatbot Area, and has since grow to be a full-fledged platform, with monetary assist from UC Berkeley, a16z, Sequoia Capital, and elsewhere.
Each consumer’s vote impacts fashions’ rankings on the scoreboard by way of the Elo system, which was initially devised to rank gamers in chess and different aggressive video games. Mannequin names stay nameless to scale back bias; solely after a vote has been submitted is the title of the mannequin revealed to the voter.
On the time of this writing, somewhat underneath 281,000 votes have been submitted on the efficiency of 9 text-to-image fashions. OpenAI’s GPT Picture 1 presently holds the highest spot, adopted by Google’s Imagen and Luma’s Photon.
Beneath the scoreboard, the web site depicts a sequence of charts that present larger element into how every mannequin’s rating is decided.
Separating the wheat from the chaff
The launch of ChatGPT in November 2022 sparked a flurry of comparable generative AI instruments as corporations vied for dominance on this new product class, which some forecast may very well be essentially the most profitable for the reason that smartphone and even the web.
In consequence, customers are confronted with a large variety of selections. Each generative AI mannequin has distinctive capabilities, pricing construction, and knowledge privateness insurance policies, making it tough to know which of them to show to. Web sites like LMArena have stepped in to offer some readability.
AI builders have lots to realize from these web sites, too, as they supply platforms for sharing precious suggestions and are primarily based on in depth testing. This results in a gentle stream of use that, in flip, helps the fashions themselves enhance. ZDNET went hands-on with dozens of AI picture turbines and recognized the highest 10, scoring them on totally different options and attributes corresponding to price, guardrails, velocity, high quality, and extra.
Get the morning’s high tales in your inbox every day with our Tech Right now e-newsletter.
