A New Contender in AI Image Generation

A groundbreaking new image generation model, “red_panda,” outperforms leading models from Midjourney, Black Forest Labs, and OpenAI on the crowdsourced Artificial Analysis benchmark.

With an impressive lead of approximately 40 Elo points over the second-best model, Black Forest Labs’ Flux1.1 Pro, red_panda dominates the text-to-image leaderboard. Artificial Analysis employs the Elo ranking system, originally designed to assess the skill levels of chess players and evaluate the performance of the models it tests.

Like the community AI benchmark Chatbot Arena, Artificial Analysis effectively ranks models through crowdsourcing. It randomly selects two contenders for image models and gives them a unique prompt. The prompt and the resulting images are then presented to users, who decide which model best captures the essence of the prompt.

While there may be some bias in the voting process, as the voters are primarily AI enthusiasts, their choices may only partially represent the broader community of generative AI users.

However, red_panda stands out as one of the top-performing models on the leaderboard, particularly in generation speed. It averages around 7 seconds to produce an image—over 100 times faster than OpenAI’s DALL-E 3.

So, where did red_panda originate? Which company developed it? And when can we expect its release? These are all excellent questions. AI labs are increasingly leveraging community benchmarks to build anticipation ahead of announcements, so we likely won’t have to wait long for the answers.