FLUX 3
FLUX 3 release timeline
| Date | Release | What it is |
|---|---|---|
| 23 Jul 2026 | FLUX 3 announced | A multimodal "real world model" family covering image, video and action |
| 4 Aug 2026 | FLUX 3 Video (part 1) | Video generation |
| 23 Sep 2026 | FLUX 3 Action | A model aimed at robotics |
| 15 Oct 2026 | FLUX 3 Image Klein | 5.6B parameters, open weights; text-to-image, multi-reference generation and editing |
FLUX 3 Image Klein
Klein is the image model most people mean by "FLUX 3" today: 5.6 billion parameters with open weights, so it runs on your own GPU or in tools such as ComfyUI once community workflows land. It handles text-to-image, generating from several reference images at once, and editing an existing image. Black Forest Labs says it leads their blind evaluations; independent comparisons are still scarce.
FLUX 3 vs GPT Image 2.6
| FLUX 3 Image Klein | GPT Image 2.6 | |
|---|---|---|
| Weights | Open | Closed |
| Run locally | Yes | No (ChatGPT or API) |
| Multi-reference editing | Yes | Yes, references persist in a chat |
| Animated GIF / SVG | No | Yes |
Compare other models: GPT Image 2.6, Qwen Image, Nano Banana Pro pricing.
Questions
What is FLUX 3?
Black Forest Labs' third model generation, announced on 23 July 2026. It spans image, video and robotics models rather than a single text-to-image model.
Is FLUX 3 open source?
FLUX 3 Image Klein, released on 15 October 2026, has open weights (5.6B parameters). Check its license before commercial use.
Can FLUX 3 edit images?
Yes. Klein supports multi-reference generation and editing as well as text-to-image.
FLUX 3 or GPT Image 2.6?
FLUX 3 Klein can run on your own hardware with open weights; GPT Image 2.6 is closed but adds animated GIF and SVG output and runs inside ChatGPT.
Source: Black Forest Labs blog (bfl.ai/blog), checked 10 Nov 2026. FLUX 3 Image pricing was not published on the pages we read.