FLUX.1

What it is

FLUX.1 is a family of text-to-image models from Black Forest Labs (founded by the original Stable Diffusion researchers). It comes in open-weight variants (schnell, dev) you can self-host and a hosted Pro variant via API, and is widely regarded as state-of-the-art for open image generation. FLUX powers many third-party tools (via Replicate, fal, Together) and can run locally on capable GPUs, making it the open flagship for developers who want control.

The open-weight flagship — state-of-the-art quality you can self-host or call via API, from the original Stable Diffusion team.

When it fits — and when it doesn’t

Known weaknesses

Each one links to the public issue, changelog or docs page it came from, where we have one.

1. 01

   Lack of official fine-tuning code and uncertainty regarding fine-tunability

   A highly reacted open issue indicates that fine-tuning code has not been released and users are unsure if the model is even fine-tunable.

   Source: GitHub issue #9 ↗
2. 02

   Missing support for Apple Core ML

   Users have requested support for transforming the model to Apple Core ML, indicating a lack of native support for Apple's on-device machine learning framework.

   Source: GitHub issue #34 ↗
3. 03

   Poor error handling and lack of progress indicators for CPU-only or non-CUDA environments

   The tool fails to provide clear warnings or progress bars when running on CPU or without CUDA, leading to confusing assertion errors and unresponsive interfaces.

   Source: GitHub issue #55 ↗
4. 04

   Ineffective environment variable handling for local model caching

   Setting environment variables to point to local model copies does not prevent the system from re-downloading files, causing unnecessary time and bandwidth consumption.

   Source: GitHub issue #23 ↗

What it does well

• State-of-the-art quality for an open model

  Widely rated top among open-weight generators; rivals closed models on many prompts

• Self-hostable open weights

  schnell (Apache-2.0) and dev variants run locally on capable GPUs — full control, no per-image fees

• Broad ecosystem and API availability

  Hosted on Replicate, fal, Together; integrates into many pipelines

Pricing

Free tier

YesOpen weights (schnell) are free to self-host

Paid from

$0.003/image (API)

Model

Usage-based

Enterprise plan

Yes

schnell open weights free to self-host; Pro via API ~$0.003-0.05/image depending on variant/host.

Pricing not independently confirmed — check the vendor’s page. Vendor pricing ↗