
Flux
by Black Forest Labs4.5
Flux comes from Black Forest Labs, founded by people behind Stable Diffusion. It ships commercial versions (Pro, Ultra) alongside an open one (Schnell), and sits at the current state of the art for realism and prompt adherence.
- Commercial use OK

- Pricing
- Usage-based API pricing
- Free tier
- No
- Vendor
- Black Forest Labs
- Commercial use
- Permitted
- Japanese UI
- Not confirmed
- Enterprise contract
- Not confirmed
What it is good for
- Developers embedding image generation through an API
- Creators who want the newest model quality
- Self-hosted pipelines using open weights
Strengths and trade-offs
Strengths
- Current state of the art on realism and prompt adherence
- Commercial and open versions side by side
- Available through Replicate, Together and other inference providers
Trade-offs
- No canvas of its own — you pair it with ComfyUI, Krea or similar
- For one-prompt finished visuals, Midjourney is quicker
Editors' review
A next-generation image model from the Stable Diffusion founders
Flux is Black Forest Labs' image model family, with commercial Pro and Ultra versions alongside the open Schnell. On realism and prompt adherence it represents the 2026 state of the art.
It is used through usage-based APIs, and weights can be downloaded from Hugging Face and similar. In practice most people call it through an inference provider such as Replicate or Together.
There is no GUI canvas, so it is combined with another interface such as ComfyUI or Krea. If you want a finished visual from a single prompt, Midjourney gets there faster.
Verdict
For developers embedding generation through an API and creators chasing the newest model. If you work in a GUI, use Midjourney.
Compiled from official documentation and public sources, with our own editorial judgement. We do not publish figures we cannot source.