On August 1, 2024, Stable Diffusion-associated creators Robin Rombach, Patrick Esser, and Andreas Blattmann announced Black Forest Labs, introduced the FLUX.1 text-to-image model family, and disclosed a $31 million seed financing round led by Andreessen Horowitz. The launch was not a single free, unrestricted Stable Diffusion replacement: FLUX.1 arrived in three versions with different access models, capabilities, and commercial rights.
Update: This is a historical account of the 2024 launch. Black Forest Labs now promotes later FLUX generations, including FLUX 3, while continuing to present FLUX.2 and FLUX.1 Kontext products. See Black Forest Labs and its current documentation for the latest offerings.
As an Amazon Associate I earn from qualifying purchases.
What Black Forest Labs announced
The August 1, 2024 announcement combined three developments:
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problems- Black Forest Labs became a new AI company.
- The company launched FLUX.1, a family of text-to-image models.
- It revealed $31 million in seed funding.
VentureBeat reported that Andreessen Horowitz led the round, with participation from investors including Brendan Iribe, Michael Ovitz, and Garry Tan. Other contemporary reports named additional backers, including Timo Aila and Vladlen Koltun. The reported figure was financing, not a company valuation; no valuation, ownership split, or investor return should be inferred from it. VentureBeat’s launch report provides the historical funding account.
#1 Best Overall
- Use scikit-learn to track an example ML project end to end
- Explore several models, including support vector machines, decision trees, random forests, and ensemble methods
- Exploit unsupervised learning techniques such as dimensionality reduction, clustering, and anomaly detection
- Dive into neural net architectures, including convolutional nets, recurrent nets, generative adversarial networks, autoencoders, diffusion models, and transformers
- Use TensorFlow and Keras to build and train neural nets for computer vision, natural language processing, generative models, and deep reinforcement learning
Who founded Black Forest Labs?
Black Forest Labs was founded by Robin Rombach, Patrick Esser, and Andreas Blattmann. They were key members of the originating Stable Diffusion effort and are often described as original Stable Diffusion creators. That description does not mean every Stable Diffusion contributor, or the entire Stability AI research organization, moved to the new company.
The founders’ prior work helped make the announcement significant. Black Forest Labs was not simply releasing another community checkpoint: it was a new, venture-backed company attempting to build frontier image models while combining paid services with selected open-weight releases.
What was FLUX.1?
FLUX.1 was a text-to-image family built around a 12-billion-parameter rectified-flow transformer. Its launch materials emphasized prompt following, image quality, typography, output diversity, and flexible aspect ratios. The family was positioned as a competitor to Stable Diffusion, Midjourney, DALL-E, and other image-generation systems.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
“Better than” is not a universal technical conclusion. Image-generation results vary with the prompt, resolution, sampler, hardware, model version, and evaluation method. Launch-era comparisons should therefore be treated as claims or results under particular conditions rather than proof that FLUX.1 won every category.
The three FLUX.1 versions
| Variant | Weights available? | Primary access | License position |
|---|---|---|---|
| FLUX.1 [pro] | No | Black Forest Labs API | Closed model governed by API and service terms |
| FLUX.1 [dev] | Yes | Hugging Face and local tooling | FLUX.1 Dev Non-Commercial License |
| FLUX.1 [schnell] | Yes | Hugging Face and local tooling | Apache 2.0, subject to applicable law and restrictions |
FLUX.1 [pro]
Pro was the premium, closed option. Users accessed it through Black Forest Labs’ API rather than downloading the model weights. That made it attractive for applications and production integrations that did not want to operate GPUs, but it also introduced API costs, vendor dependency, service terms, and availability considerations.
Rank #2
FLUX.1 [dev]
Dev supplied open weights for research, experimentation, and development. It used the FLUX.1 Dev Non-Commercial License, not Apache 2.0. A company should not assume that downloading the weights permits commercial client work, a paid SaaS product, a hosted inference service, or unrestricted redistribution. Review the current terms and obtain appropriate commercial permission where required. The official Dev model card also lists prohibited uses and limitations.
FLUX.1 [schnell]
Schnell was the faster, distilled version, designed to produce images in one to four steps. It was released under Apache 2.0, making it the simpler choice for many local commercial prototypes, provided users comply with the license, applicable law, and other relevant restrictions. Apache 2.0 does not automatically resolve copyright, privacy, publicity-rights, or safety questions surrounding generated content. See the Schnell model card.
Is FLUX.1 open source?
That depends on the version—and “open source” is too broad a label for the whole family.
- Pro: closed and accessed through an API.
- Dev: open weights, but distributed under a non-commercial license with additional restrictions.
- Schnell: open weights under Apache 2.0.
The important distinction is between open weights and commercially permissive licensing. Open weights let a user obtain and run model files; they do not necessarily grant permission to build a commercial service, fine-tune and redistribute a derivative, or use the model for every business purpose. The official repository links to the relevant model and license information.
What technical approach set FLUX.1 apart?
FLUX.1 used rectified flow, also called flow matching, with a transformer-based architecture. In simplified terms, traditional diffusion models learn to reverse a gradual noising process. Rectified-flow systems learn a more direct transformation from noise toward an image. Distillation can then reduce the number of sampling steps, as it did for Schnell.
That architecture can support efficient generation, but it does not guarantee identical speed everywhere. Runtime depends on the model variant, GPU or accelerator, precision, resolution, implementation, memory management, and sampling settings.
Free tools Windows power users keep installed
One-click scans. No signup required.
How could users try FLUX.1?
Hosted API or browser tools
Users who do not want to manage model files and GPUs could use Black Forest Labs’ hosted products or a third-party inference provider. This is generally the easiest route for applications and automation. Before committing to a provider, check the current model list, pricing, data-handling policy, output terms, rate limits, and commercial-use conditions. Black Forest Labs’ current API documentation is the appropriate starting point.
Local installation
The original repository supplied an installation path based on Python 3.10:
cd "$HOME"
git clone https://github.com/black-forest-labs/flux
cd "$HOME/flux"
python3.10 -m venv .venv
source .venv/bin/activate
pip install -e ".[all]"That command installs the code; it does not guarantee a successful run. Local users also need a compatible accelerator, sufficient memory or CPU offloading, suitable PyTorch and driver versions, storage for large downloads, and any required Hugging Face access approvals. CUDA out-of-memory errors, slow offloading, unsupported precision, and long startup times are normal troubleshooting risks for a model of this scale.
Diffusers and workflow tools
The model cards document a Hugging Face Diffusers workflow beginning with:
Recommended Free Tools
Rank #4
pip install -U diffusers transformers accelerateSchnell’s example uses a CUDA-oriented pipeline and torch.bfloat16; the card notes that Apple users can switch to an MPS device. Dev’s example uses CPU offloading to reduce VRAM pressure. Diffusers and tools such as ComfyUI made FLUX.1 more accessible, but third-party workflows can differ in memory use, supported features, and maintenance.
Practical risks for creators and businesses
Commercial-license mistakes
The largest business risk was treating Dev and Schnell as interchangeable. A designer using Schnell under Apache 2.0 faced a different licensing situation from a company deploying Dev under its non-commercial terms. Businesses should identify the exact checkpoint, review the license, document any commercial permission, and separately consider rights in training data, prompts, people, logos, and generated outputs.
Hardware and operating costs
A 12-billion-parameter model can impose meaningful GPU, storage, electricity, and engineering costs. A local workflow may improve privacy and customization while requiring capital or recurring infrastructure spending. A hosted API reduces operational work but creates usage-based costs and dependence on a provider. The cheaper-looking route is not always cheaper at production volume.
Access and model limitations
Hugging Face may require users to accept conditions and provide contact information before accessing model files. The Dev model card warns that the model can fail to match prompts, is not intended to provide factual information, may amplify societal biases, and is strongly influenced by prompting style. Generated images should be reviewed rather than treated as authoritative or automatically safe.
Why the funding and launch mattered
The announcement mattered for more than its headline funding number. It brought prominent Stable Diffusion-associated researchers together in a new company, offered the open-weight ecosystem a major alternative during uncertainty around Stability AI, and tested a hybrid business model:
Best Value
- a closed premium model sold through an API;
- an open-weight development model with non-commercial restrictions; and
- a faster open-weight model under Apache 2.0.
That structure illustrated the tension at the center of modern AI: developers want inspectable, locally runnable models, while companies need revenue, infrastructure funding, and control over commercial deployments. FLUX.1 was therefore not “Stable Diffusion 4.” It was a new architecture, company, product family, and licensing strategy.
How FLUX.1 compared with alternatives
| Option | Strength | Trade-off |
|---|---|---|
| Stable Diffusion-family models | Broad tooling, checkpoints, and community ecosystem | Quality, licensing, and hardware needs vary by release |
| Midjourney | Polished consumer creative workflow | Not designed around self-hosted weights or deep customization |
| DALL-E and similar hosted tools | Convenient managed access | Generally closed and not locally deployable |
| Third-party FLUX hosts | No GPU operation and access to multiple versions | Extra vendor, billing layer, queue, and terms-of-service dependency |
For casual creation, a hosted interface is usually simplest. For developer prototyping, compare a direct Black Forest Labs API with a third-party host. For local experimentation, Schnell was the most straightforward original choice because of its speed-oriented design and Apache 2.0 license. For commercial SaaS or client work, do not rely on Dev’s default terms without obtaining the necessary rights.
What happened after the 2024 launch?
Black Forest Labs’ current website no longer presents FLUX.1 as its leading family. As of September 2026, it promotes FLUX 3 and also lists later FLUX.2 and FLUX.1 Kontext product lines. Current customers should consult the company’s licensing page, pricing page, and API documentation rather than assuming that 2024 model terms or availability remain unchanged.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →The original launch nevertheless remains important: it established Black Forest Labs as a funded competitor and made FLUX.1 a notable case study in how AI companies can combine proprietary services with selectively open model releases.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
