FLUX1.1 [pro]

What it is

FLUX1.1 [pro] is a text-to-image model released by Black Forest Labs (BFL) on 2 October 2024, the same day the company's BFL API became generally available after a beta period.^[1] It is a closed model, offered only through an API rather than as something users can download and run themselves.^[2] A user sends a text prompt and the model returns an image.

Where it came from

The model is the successor to FLUX.1 [pro]. BFL said FLUX1.1 [pro] generates six times faster than FLUX.1 [pro], and that it had also updated FLUX.1 [pro] itself so that the older model ran about twice as fast.^[1] Before release, the new model was tested in the Artificial Analysis image arena under the codename "blueberry", where it reached the highest overall Elo score on that leaderboard.^[1] Elo is a rating scheme borrowed from chess that turns a series of head-to-head comparisons into a single number, so a top Elo score on an arena leaderboard means the model came out ahead in more of those comparisons than its rivals.

Speed at high resolution

BFL stressed resolution as much as speed, saying FLUX1.1 [pro] was natively set up for fast ultra-high-resolution generation and promising images of up to 2k through the API without sacrificing prompt following, which is how closely the output matches what the prompt asked for.^[4]

Access, price and controls

On the BFL API, the model was priced at 4 cents per image, against 5 cents for FLUX.1 [pro].^[1] A published price list put FLUX.1 [dev] at 2.5 cents per image, FLUX.1 [pro] at 5 cents and FLUX1.1 [pro] at 4 cents.^[4] Outside BFL's own API, the model was also offered through Together.ai, Replicate, fal.ai and Freepik.^[1] The API let developers choose which model to call, set the image resolution, and configure content moderation.^[4]

Modes

On 6 November 2024, BFL added Ultra and Raw modes to FLUX1.1 [pro], giving the model generation options beyond its default mode.^[3]

Where it sits in the line

For BFL, the launch paired a faster flagship model with the public opening of its own API, so the company was selling access to its models directly rather than only publishing them.^[1] In the published price list, FLUX1.1 [pro] was cheaper than FLUX.1 [pro] and dearer than FLUX.1 [dev], while being the faster of the two [pro] models by BFL's own account.^[1][4] Because it is closed and API-only, users cannot run it locally, and usage is metered per generated image.^[2][4] Distribution through Together.ai, Replicate, fal.ai and Freepik meant developers who did not integrate with BFL directly could still call it.^[1]