Black Forest Labs Expands Access to Frontier Visual Intelligence With Open Weights Release of FLUX 3 Image Klein
Building on the open weights release of FLUX 3 Action, the company continues to make its models available for
Press Release Disclaimer: This is a press release distributed through the XPR Media network. It has not been independently verified by our newsroom.
FREIBURG, Germany, Oct. 15, 2026 (GLOBE NEWSWIRE) — Black Forest Labs (BFL), the frontier AI research lab building visual intelligence, today announced the open weights release of FLUX 3 Image Klein, its most capable open weights image model to date. FLUX 3 Image Klein provides the quality of a large frontier model with the customization and speed of open weights, runs on consumer GPUs and laptops, and retains features from the larger FLUX 3 Image model such as native 4K generation, bounding boxes, and pixel perfect editing.
The release includes a base model for fine-tuning, a distilled model for low-latency generation, and quantized FP8, NVFP4, and INT8 versions. Following September’s open weights release of FLUX 3 Action for robotics, today’s release brings the company’s latest research to core creative use cases, giving developers and businesses greater freedom to build, customize, and deploy visual AI.
Making its research and technology available for others to build on is one of Black Forest Labs’ founding principles. BFL has consistently released its research findings alongside open weight versions of its industry-leading FLUX models, enabling others to build on its advances. So far in 2026, the team has published its Self-Flow research and released open weights for the FLUX.2 [klein] family in January and FLUX 3 Action in September. Self-Flow improved generation across images, video, and audio, while helping models learn representations that can also support robot control. This research led to BFL’s most capable model to date, FLUX 3, which brings multimodal generation and action prediction into a shared architecture. Today’s release opens access to the creative capabilities of the same FLUX 3 backbone, with further open weights releases from the FLUX 3 family to follow.
“By sharing our research openly, we created a global community of developers and creators who are helping to advance the field together,” said Robin Rombach, co-founder and CEO of BFL. “FLUX 3 allows that community to adapt the same foundation to very different needs, from creative generation to robotics. Giving developers the models, research, and tools to do that is key to advancing visual intelligence.”
Opening Access to Visual Intelligence
By releasing open weights, BFL puts its models in the hands of developers and enterprises to build on, customize, secure, and deploy independently. Teams with data sovereignty or local deployment requirements can evaluate and adapt the models on their own infrastructure. Each open weights release includes tools and support for adaptation alongside safeguards to reduce misuse.
FLUX 3 Image Klein is BFL’s most capable open weights image model to date. In BFL’s internal evaluations, FLUX 3 Image Klein scores higher than current leading open weights image models in text-to-image, multi-reference generation, and image editing, while costing less through the company’s API. At 5.6 billion parameters, it runs on consumer GPUs and laptops with as little as 8 GB of VRAM, and draws on FLUX 3’s shared learning across images, video, and audio. Images contribute spatial detail; video adds motion and physical dynamics; and audio provides information about events and interactions. Together, these signals help the model learn how the visual world behaves.
Leads in quality at a smaller model size.
Extends the quality and speed frontier.
The same research foundation extends to physical AI. Last month, BFL released FLUX 3 Action, a seven-billion-parameter open weight model for robotics that ranked No. 1 on Physics-IQ Verified and NVIDIA’s RoboLab-120 benchmark, which evaluates robot models across 120 simulated manipulation tasks. By releasing the weights and a full technical report, BFL made those capabilities available for researchers and robotics teams to train, adapt, and build on.
Accessing and Customizing FLUX 3 Image Klein
The open weights release of FLUX 3 Image Klein builds on BFL’s roots in image generation. Founded on the belief that visual intelligence starts with a pixel, BFL has been developing frontier, production-ready image models since 2024 that power creative tools across platforms including Adobe, Canva, Meta, and Microsoft. Open weights image models built by BFL’s founding team, from the original Stable Diffusion to FLUX and their community derivatives, remain among the most widely adopted, with more than 640 million downloads on Hugging Face.
FLUX 3 Image Klein gives developers a foundation they can adapt to their own applications. It supports text-to-image generation, image editing, and generation using multiple reference images. BFL is releasing both a base model for fine-tuning and research, and a step distilled version that generates images in fewer sampling steps, with open weights allowing teams to customize and run the models on their own infrastructure. The distilled version is also available through BFL’s API, where it runs with BFL’s prompt enhancement and pixel-level editing pipeline for production workloads.
Support for Ostris’s AI Toolkit at launch will help developers fine-tune the model using their own examples. The release will also be supported by developer tools and platforms including ComfyUI, Diffusers from Hugging Face, vLLM-Omni, and SGLang Diffusion, making it easier to customize and deploy. FLUX 3 Image Klein was trained on NVIDIA GB200 GPUs NVL72 and optimized for NVIDIA’s accelerated computing platform. Developers can fine-tune it with NVIDIA NeMo AutoModel and run it locally on NVIDIA GeForce RTX GPUs with 8 GB or more of VRAM.
Available under the FLUX Kommunity License, which allows free commercial use for individuals and companies with annual revenue under $5M, alongside research and non-commercial use. Larger organizations can deploy commercially through BFL’s self-hosted licensing options or API. This flexibility is particularly valuable for enterprises seeking greater control over model behavior, data handling and deployment.
Pairing Open Weights with Safety Standards
In an effort to make AI technology both open and safe, BFL implements multiple safeguards throughout the model development lifecycle to combat misuse. These include dataset filters during pre-training to limit exposure to harmful content, and targeted fine-tuning during post-training to suppress harmful behaviors, making it progressively harder to elicit harmful content. For FLUX 3 Image Klein, BFL engaged its external safety partner Cinder to evaluate both model checkpoints ahead of release, demonstrating fewer vulnerabilities than other popular open weight models. More on BFL’s safety methodology is available here.
About Black Forest Labs
Black Forest Labs is a global frontier AI research lab building the foundation layer for visual intelligence. Founded by the researchers who pioneered latent diffusion, Stable Diffusion and the FLUX model family, the small team of 100 advances research and ships open, production-ready models from Freiburg, Germany and San Francisco, USA. Valued at $3.25 billion, the company has raised more than $450 million from leading venture and strategic investors including a16z, AMP, Salesforce Ventures, NVIDIA, General Catalyst, Adobe Ventures, Figma Ventures, Canva and Deutsche Telekom’s T.Capital. Learn more at bfl.ai.
Media Contact
press@blackforestlabs.ai
Photos accompanying this announcement are available at
https://www.globenewswire.com/NewsRoom/AttachmentNg/36ec59a1-648e-4e1d-b944-c2108e8992d9
https://www.globenewswire.com/NewsRoom/AttachmentNg/953ac532-ba20-4cfe-86ad-8ae6497349ba
