Introduction

Great news! Black Forest Labs has introduced Flux.1, a new revolutionary image generation AI model. We have studied and tested the new model, which is making waves on the internet, climbing various rankings, and receiving positive feedback from users. In this article, we will tell you everything you need to know about it: who made it, what it can do, image examples, how to use it (and if you can use it for free), and links to useful resources. Let's dive in step by step. Let's go!

Who Made It?

The core team consists of the creators of another well-known model, Stable Diffusion, who left Stability-ai in March 2024. As many expected, they left and started their own company!

It's worth noting that the official announcement of the new company was made only on August 1st. The new team managed to create and train a new engine in just a few months, which sounds incredible considering the final product's quality.

Capabilities of Flux.1

FLUX.1 is a Text2image model with a whopping 12B parameters! It is currently the state-of-the-art model available to the public. It seems FLUX.1 was made following the recipe of Stable Diffusion 3, as it has a very similar architecture (DiT with two streams – text and image) and is also based on Flow Matching.

The model is breaking the charts on the text-to-image Arena. Even the trimmed-down open-source FLUX.1 Schnell, which can be used for anything due to its Apache 2.0 license, lags in quality only behind Midjourney v6 and SD3 Large.

Model Strengths

FLUX has a lot of strong points, which is why it got such positive feedback right after its launch. Here are the key ones:

• High-Quality Image Generation: Flux.1 is great at creating hyperrealistic and detailed images, especially when it comes to complex features like human hands, which have been tough for other AI models. Its advanced training helps it make anatomically correct and visually appealing images, setting a new standard in text-to-image generation.

• Prompt Understanding: The model accurately follows instructions, even with long prompts full of details and clarifications, showing exactly what the user wants.

• Speed and Efficiency: The model comes in different versions to suit various needs, including Flux.1 Schnell, which is designed for speed.

• Flexible Configuration: Flux allows you to adjust a wide range of generation settings: set different image sizes and aspect ratios, the number of inference steps, safety tolerance levels, and more.

• Open-Source Accessibility: The model is available in open-weight formats, making it easy for developers and researchers to experiment and build on its features.

Available Versions

The new model comes in three versions:

1. FLUX.1 [pro]

The state-of-the-art model with 12B parameters. It offers good image detail and prompt adherence, with various styles.

2. FLUX.1 [dev]

This is a distilled version of FLUX.1 [pro] using Guidance Distillation, which is about twice as fast as the original and delivers almost the same quality.

3. FLUX.1 [schnell]

This is the fast version. It appears to be distilled using LADD (Latent Adversarial Diffusion Distillation), working in a small number of steps (from 1 to 12) but with slightly lower quality.

Getting Access to FLUX.1

Currently, the [pro] and [dev] models are under commercial use, but some services support playgrounds where you can use Flux 1 pro for free a limited number of times. The [schnell] model is open-source and available on more platforms.

Replicate (Access to Pro, Dev, Schnell)

The service allows you to use the Schnell version for free on their website and provides access to all three versions via API. You can check the pricing for the Dev and Pro versions here.

To get access:

1. Visit the website: https://replicate.com/black-forest-labs

2. Choose the model you want to use.

3. If you picked Schnell, you're already done. For paid options, you have to set up billing first by clicking Manage Billing button.

ChatLabs (Access to Flux.1 Pro)

1. Go to the official ChatLabs website: https://chatlabs.pro

2. Sign up using your email.

3. Activate the ChatLabs Pro subscription. Using Flux.1 Pro requires a paid subscription.

4. Select Flux.1 Pro in the Plugins section.

ChatLabs provides access not only to the Pro model from Black Forest Labs but also to all the best models on the market. This includes both text and image models: DALL-E 3, Stable Diffusion 3, GPT-4.5 Turbo, GPT-4o, Claude 3.5, Llama 3.1, and many others.

Fal.ai

This service provides access to the models through a paid API (pricing). Additionally, there is a Playground where you can generate around 20 images on each of the Pro and Dev models from a single account completely free.

How to use Playground chats:

1. If you want to use:

   • Pro, go to: https://fal.ai/models/fal-ai/flux-pro

   • Dev, go to: https://fal.ai/models/fal-ai/flux/dev/playground

   • Shcnell, go to: https://fal.ai/models/fal-ai/flux/schnell/playground

2. Sign up with your email or GitHub account.

3. Start making beautiful pictures!

Basedlabs (Access to Flux.1 Schnell)

This service also allows you to use the fastest model for free.

1. Head to the official website: https://www.basedlabs.ai and sign up with your email

2. Choose FLUX in the model picker menu.

3. Start creating images!

Image Examples

The quality of the generated images is truly impressive. Realism, adherence to prompts, depiction of human hands, handling of light, and attention to detail – everything is top-notch.

Conclusion

The team at Black Forest Labs has done an excellent job, giving us a new, very powerful, and high-quality t2i model. Benchmarks, user reviews, and authoritative rankings only confirm this. Will the new model become the "killer" of Midjourney and other competitors? Time will tell, but it has a good chance of becoming the absolute leader among image generators for a relatively long time.

As always, we made the new image generation model available to Pro users of our ChatLabs platform immediately after its release, striving to provide access to the best models on the market as quickly as possible.

Useful Links

• Official website: https://blackforestlabs.ai

• Flux 1 weights on GitHub: https://github.com/black-forest-labs/flux

• Text-to-Image Arena by Artificial Analysis: https://artificialanalysis.ai/text-to-image/arena

• The models on HuggingFace: https://huggingface.co/black-forest-labs