Try “Veo”, “voiceover”, “thumbnail” or “Descript” · Esc to close

FLUX.2 🧠 Model Open source

Black Forest Labs' FLUX.2 family: hosted [max]/[pro]/[flex] plus open-weight [dev] and [klein]

Image models · Open source ★ 2.7k · Apache-2.0 · updated 2026-03-12

8.6editor score
Visit FLUX.2 ↗
GitHub stars
2.7k
Stars this week
–
Forks
193
Licence
Apache-2.0
Last push
2026-03-12
Maintainer
Black Forest Labs

Where you can use FLUX.2

Apps in this directory that let you generate with FLUX.2, from their own documentation.

See every app × model on the model map →

Good for

About FLUX.2

What it is

FLUX is the image model family from Black Forest Labs (BFL), the Freiburg team behind the original Stable Diffusion research. FLUX.2 launched on 25 November 2025 with [pro], [flex] and the open-weight 32B [dev]; the fast [klein] models (4B and 9B) followed on 15 January 2026, and a top [max] tier sits above [pro]. In 2026 BFL also launched FLUX 3 as a video-and-audio model, but FLUX.2 remains its image line.

Key features

  • Multi-reference generation with up to 10 images for consistent products, characters and styles
  • Editing at up to 4 megapixels while preserving detail
  • Readable typography for infographics, memes and UI mockups
  • [klein] generates and edits in under a second and runs on-device
  • LoRA training on the base weights for your own brand style

Where you can use it

BFL's API and Playground, plus hosts such as fal, Replicate and Together AI. Creative apps that document FLUX models: Krea (FLUX.2 Flex, Klein, Max and Pro), Adobe Firefly (FLUX.2 [pro] partner model), Picsart (Flux 2 Pro), Higgsfield, Leonardo.ai, and BFL names Canva and Freepik as FLUX.2 partners. Locally, ComfyUI and InvokeAI support FLUX.2 [dev] and [klein].

Pricing and rights

BFL's API is pay-as-you-go with credits at $0.01 each; FLUX.2 images are priced per megapixel, from about $0.014 on [klein] to $0.07 on [max]. There is no consumer subscription. FLUX.2 [klein] 4B is Apache 2.0 (commercial use allowed). FLUX.2 [dev] and [klein] 9B use non-commercial licences; BFL sells self-hosting licences for commercial use. Outputs made through partner apps follow each app's terms.

Who it is for

Designers and studios who want top image quality in the tools they already use, and technical creators who run models locally in ComfyUI.

Verdict

FLUX.2 is the most widely integrated image family and the only top-tier one with usable open weights. Read the licence carefully: only the small [klein] 4B model is free for commercial self-hosting.

image-generation open-weights black-forest-labs comfyui editing

Pros

  • FLUX.2 [klein] 4B is Apache 2.0, free for commercial use and runs on a laptop
  • Up to 10 reference images for consistent products and characters
  • Edits at up to 4 megapixels with legible typography
  • Offered in more creative apps than almost any other model family

Cons

  • FLUX.2 [dev] and [klein] 9B weights are non-commercial unless you buy a licence
  • No consumer subscription from BFL; you use it through partners or the pay-as-you-go API
  • BFL's 2026 focus has shifted to FLUX 3 video; no FLUX 3 image model yet
  • Self-hosting [dev] (32B) needs a high-VRAM GPU or quantisation

Similar AI models

All image models →

GPT Image 2.5 🧠 ModelFreemium

OpenAI's image model behind ChatGPT Images, with 4K output and up to 16 reference images

OpenAI · 2026-09

9.0 Visit ↗

Nano Banana 2 🧠 ModelFreemium

Google's Gemini image models (Nano Banana 2, Pro and 2 Lite) for generation and editing up to 4K

Google DeepMind · 2026-02

9.0 Visit ↗

Midjourney V8.2 🧠 ModelPaid

Midjourney's own image model, known for aesthetics and personalization, now in V8.2

Midjourney · 2026-07

8.5 Visit ↗

Seedream 5.0 Pro 🧠 ModelFreemium

ByteDance's Seedream image models, used in CapCut, Dreamina and many creative apps

ByteDance Seed · 2026-07

8.3 Visit ↗

Ideogram 4.0 🧠 ModelOpen source

Ideogram's text-rendering image model with JSON prompting, now also as non-commercial open weights

Ideogram · 2026-06 · open weights

8.0 Visit ↗

Qwen-Image 3.0 🧠 ModelOpen source

Alibaba's Qwen image models, known for dense multilingual text rendering, with open-weight versions

Alibaba (Qwen team) · 2026-09 · open weights

Popular searches