FLUX Timeline: Black Forest Labs Model Releases & History

Black Forest Labs launched FLUX.1 in August 2024 as a text-to-image model family. FLUX1.1 improved the original generation line, FLUX.1 Kontext introduced text-and-image editing, FLUX.2 expanded multi-reference generation and editing, and FLUX 3 moved beyond still images with video, native audio, and action prediction.

This timeline follows that progression from FLUX.1 on August 1, 2024 through the current FLUX 3 rollout. It records the major model releases and changes in public availability, including early access, API launches, open-weight releases, and deprecations.

Last updated: October 7, 2026.

FLUX model timeline

2026

Date	Milestone
October 1, 2026	FLUX 3 Image launches. Image generation and editing become available through the BFL API, with bounding-box control for layouts and targeted edits.
September 23, 2026	FLUX 3 Action releases open weights. The 7-billion-parameter world-action model adapts the FLUX 3 backbone for joint video and action prediction in robotics.
August 4, 2026	FLUX 3 Video opens public API access. The initial text-to-video and image-to-video release generates clips up to 20 seconds long with synchronized native audio.
July 23, 2026	FLUX 3 enters early access. The new multimodal foundation model is trained on images, video, and audio. Video, image, action, and open-weight releases follow their own rollout schedules.
January 15, 2026	FLUX.2 [klein] launches. Compact 4B and 9B models unify image generation and editing, with distilled and Base variants. The 4B models use Apache 2.0, and the 9B models use the FLUX Non-Commercial License.

2025

Date	Milestone
December 16, 2025	FLUX.2 [max] launches. The highest-end FLUX.2 variant adds grounding search for image generation informed by current web information.
November 25, 2025	FLUX.2 launches. The second generation unifies image generation and multi-reference editing. Hosted [pro] and [flex] variants launch alongside the 32-billion-parameter open-weight [dev] model.
October 31, 2025	Original FLUX Pro API deprecations take effect. BFL deprecates the FLUX Pro 1.0, Depth, and Canny endpoints and the FLUX.1-era Finetuning API.
July 31, 2025	FLUX.1 Krea [dev] releases. Developed with Krea AI, the open-weight text-to-image model focuses on a photographic aesthetic with less oversaturated texture.
June 26, 2025	FLUX.1 Kontext [dev] releases public weights. The 12-billion-parameter editing model moves from private beta to a public open-weight release under the FLUX.1 Non-Commercial License.
May 29, 2025	FLUX.1 Kontext launches. Hosted [pro] and [max] models introduce text-and-image generation and iterative image editing. Kontext [dev] begins in private beta, and the BFL Playground launches.

2024

Date	Milestone
November 21, 2024	FLUX.1 Tools releases. Fill adds inpainting and outpainting, Depth and Canny add structural guidance, and Redux adds image variations.
November 6, 2024	FLUX1.1 [pro] Ultra and Raw modes launch. Ultra supports images up to 4 megapixels, and Raw introduces an aesthetic focused on candid photography.
October 2, 2024	FLUX1.1 [pro] launches with the public beta BFL API. The model improves generation speed, image quality, and prompt adherence over the original FLUX.1 [pro].
August 1, 2024	FLUX.1 launches. The original text-to-image family includes hosted [pro], open-weight [dev] for non-commercial use, and Apache 2.0-licensed [schnell].

How the FLUX model family evolved

Model family	Main change
FLUX.1	Black Forest Labs’ original text-to-image family, released as [pro], [dev], and [schnell]. FLUX.1 Tools later introduced Fill, Depth, Canny, and Redux for image modification and structural control.
FLUX1.1 [pro]	A faster successor to FLUX.1 [pro] with improved image quality and prompt adherence. Ultra later increased output resolution, and Raw introduced a more natural photographic style.
FLUX.1 Kontext	Expanded FLUX.1 from text-to-image generation into generation and editing with text and image inputs. Kontext [dev] later became the open-weight editing model in this branch.
FLUX.2	Unified image generation and editing with multi-reference input, higher-resolution editing, stronger typography, and hosted plus open-weight variants.
FLUX 3	Moved the family into a multimodal architecture trained on images, video, and audio, followed by separate Image, Video, Action, and planned Dev releases.

Current FLUX 3 rollout status

FLUX 3 branch	Current status
FLUX 3 Image	Available through the BFL API. Black Forest Labs also provides commercial weights licensing for self-hosted deployment.
FLUX 3 Video	Generally available through the BFL API and selected partners for text-to-video and image-to-video generation with native audio.
FLUX 3 Action	Released as a 7B open-weight world-action model for joint video and action prediction, with weights and code available publicly.
FLUX 3 Dev	Announced as the open-weight multimodal backbone. A public release date has not been confirmed as of October 8, 2026.

FAQs

What do the brackets in FLUX model names mean?

Labels such as [pro], [dev], [max], and [klein] identify variants within a generation or model branch. They do not create a new generation number. FLUX.1 Kontext [dev], for example, belongs to the Kontext editing branch, and FLUX.1 [dev] belongs to the original FLUX.1 text-to-image family.

Why can one FLUX model have more than one release date?

Black Forest Labs can announce a model before its API access, public weights, or general availability arrive. FLUX 3 is a clear example: the family entered early access on July 23, 2026, followed by separate public releases for Video, Action, and Image.

Does open-weight FLUX mean unrestricted commercial use?

No. Open weights describe access to model parameters. The license determines permitted use. FLUX.1 [schnell] and FLUX.2 [klein] 4B use Apache 2.0, while FLUX [dev] models and FLUX.2 [klein] 9B use non-commercial licenses unless separate commercial licensing applies.