Announcing FLUX1.1 [pro] by BFL and FLUX [dev] speed improvements

fal is proud to partner with Black Forest Labs to offer both the new FLUX.1.1 [pro] (aka 🫐) and the 2x accelerated FLUX.1 [dev] models.

To play around with the new FLUX.1.1 [pro] now, check out the demo page here on fal.

Codename "blueberry 🫐"

The FLUX1.1 [pro] model, codenamed "blueberry," was introduced in the Artificial Analysis image arena (https://artificialanalysis.ai/text-to-image/arena) last week. It currently boasts the highest overall Elo score among all image models, while at the same time being one of the fastest.

Better LoRA training performance, no compromise on quality

In addition to these new models from BFL, we are also happy to announce new speed records on FLUX.1 [dev] LoRA trainer (<2 minute per 1000 step). This is currently the fastest LoRA trainer by a wide margin. 

2x Faster FLUX.1 [dev]

Today we are also launching a new endpoint for the FLUX.1 [dev] model. This new endpoint called Reference API is 2x faster than our already very fast FLUX.1 [dev] endpoint without compromising quality (~1.5second/image at 30 steps). Try it here! It is currently in an experimental version, but we are hoping to graduate it to prod soon and let everyone use it!