A mysterious new image generation model emerges.

A new image generation model, known as "red_panda," has outperformed other notable models such as those from Midjourney and Black Forest Labs in the crowdsourced benchmark by Artificial Analysis. Currently, red_panda leads the text-to-image generation performance list with an Elo score 40 points above the second-best ranked model, Black Forest Labs' Flux1.1 Pro. This Elo-based ranking system—originally created to evaluate chess skills—is used by Artificial Analysis to compare the performance of different models.

Like other community benchmarks such as Chatbot Arena, Artificial Analysis evaluates image models through a crowdsourcing process. In this methodology, two models are randomly selected and presented with a specific prompt, after which the results are displayed and users vote to determine which image best represents the prompt.

It is important to note that the voting process may be biased, as most voters are AI enthusiasts, which might not reflect the preferences of the entire community of generative AI users. Despite this, red_panda also stands out in generation speed, requiring an average of only 7 seconds to create an image, which is over 100 times faster than OpenAI's DALL-E 3.

Many questions remain about the origin of red_panda, the identity of the company behind its development, and when its launch is expected. As AI labs begin to use these community benchmarks to build anticipation before an announcement, it may not be long before more details about this innovative model are revealed.