The AI ​​of the mysterious illustrator red_panda was introduced with stunning performance

The competition for the production of advanced artificial intelligence is moving to new stages. Recently, an unknown model called “red_panda”, which is used to generate images, has been able to defeat the advanced models of Midjourney, OpenAI and Black Forest Labs. The images produced by this model look so real that it is really difficult to distinguish them.

This artificial intelligence model has been seen in the Artificial Analysis benchmark and it is about 40 Elo points ahead of the next best model, Flux1.1 Pro made by Black Forest Labs. The Artificial Analysis benchmark uses a rating measure called Elo to compare the performance of different tested models. This measure was originally developed to calculate the relative skill level of chess players.

The images produced by red_panda are very close to real images

The Artificial Analysis benchmark classifies different artificial intelligence models through crowdsourcing. For image models, this center first selects 2 random models and then sends them a text request. Ultimately, it is up to users to decide which of these models produces more realistic images.

Some users may vote for their favorite models out of prejudice. Most voters are AI enthusiasts and their choices may not reflect the preferences of the wider AI user community.

The red_panda model is also one of the best tested models in terms of photo production speed. This model needs an average of 7 seconds to produce each image; Almost twice as fast as OpenAI’s DALL-E 3 model.

Despite the extraordinary performance of the red_panda model, it is not yet clear which company made this model and when it will be publicly available to users. Artificial intelligence labs usually use such benchmarks to measure the performance of their models before they are officially released to the market. We will probably hear more details about the red_panda model soon.