Published: October 4, 2026
Image Model
Flux 3 Image is Black Forest Labs' FLUX 3 image model, covering text-to-image and editing in one. It's strongest at layout control, accurate text rendering and native 4K output — a clear step up from Flux 2 Pro and Kontext, and a contender for Nano Banana Pro and GPT Image 2 on text-heavy work.
Variants at a Glance
Type |
Modalities |
Max Res. |
Talking Points |
|
|---|---|---|---|---|
| Variant | ||||
| Text-to-Image | Image | Text-to-Image | 4K | Posters, packaging, on-image text. |
| Image Edit | Image | Image-to-Image | 4K | Local edits, or merge up to 10 references. |
Key Features
Generation
- Text rendering: Headlines and signage spelled correctly, in multiple languages.
- Layout control: Holds multi-subject prompts and placement instructions.
Editing
- Local edits: Change one object or region; the rest of the frame stays.
- Multi-reference: Combine subject, product, background and texture references.
Technical Capabilities
Details |
|
|---|---|
| Quality / Resolution | 0.75K, 1K, 2K, 4K |
| Aspect Ratios | 21:9, 2:1, 16:9, 3:2, 7:5, 4:3, 5:4, 1:1, 4:5, 3:4, 5:7, 2:3, 9:16, 1:2, Auto |
| Images per Generation | 1–4 |
Limitations
- 4K can take several minutes per image. Iterate at lower tiers.
- In edits, Auto aspect ratio follows the first reference, so reference order matters.
- References must be at least 256px per side and no larger than 4 MP.
Prompting Tips
- For edits, name only the change ("turn the white heart red") and refer to references by number ("image 1", "image 2").
- Put on-image text in quotes and say where it sits. Hex codes work for exact colors.