Black Forest Labs

TL;DR. Black Forest Labs is a Freiburg and San Francisco AI lab that builds FLUX image generation and editing models. Founders include Stable Diffusion creators Robin Rombach, Andreas Blattmann, and Patrick Esser. FLUX ships through an application programming interface (API), downloadable model weights, and enterprise deployments. Model weights let teams run FLUX on their own hardware. It suits developers, creative teams, and enterprises building production visual workflows.

At a Glance

The Stable Diffusion research team turned its visual-model work into FLUX, an AI lab based in Freiburg and San Francisco. Black Forest Labs launched in August 2024 to advance generative models for images and video. It now describes its focus as foundational infrastructure for visual intelligence. FLUX models generate and edit images, with open-weight releases, API access, and enterprise offerings.

Founders Robin Rombach, Andreas Blattmann, and Patrick Esser previously helped create Stable Diffusion. At launch, BFL announced a $31 million seed round led by Andreessen Horowitz. It also named angel investors and follow-on backing from General Catalyst and MätchVC.

On December 1, 2025, it announced a $300 million Series B at a $3.25 billion post-money valuation. Salesforce Ventures and Anjney Midha’s AMP co-led the round. The current careers page lists $450 million in total funding raised, a figure also reported by Crunchbase News. Other Series B participants included NVIDIA, Northzone, Creandum, Earlybird VC, and General Catalyst.

Freiburg is the company’s headquarters, and it also has a San Francisco office. BFL does not publish an employee count on its About or careers pages. It describes itself as a small research lab. Key links: website, About, API documentation, GitHub inference repository, and blog.

Founders & Leadership

Ten people appear on investor Earlybird’s founder list, while Black Forest Labs identifies Robin Rombach as co-founder and CEO. Public leadership information centers on the founding researchers, not a broad executive roster.

Earlybird lists ten founders. Salesforce Ventures names Rombach, Blattmann, and Esser as founders, and identifies Rombach as CEO. Black Forest Labs credits its team with work on VQGAN, latent diffusion, Stable Diffusion, and other diffusion models. The company’s August 2024 launch post also named Michael Ovitz and Professor Matthias Bethge to its advisory board.

The company describes its research lab as small and fast-growing, with offices in Freiburg and San Francisco. Its careers page lists four values: “Obsessed,” “Low ego,” “Bold,” and “Kind.” It also emphasizes curiosity, shared credit, and empathy. Reviewed official pages do not provide a verified employee count or a fuller current C-suite roster.

How It Makes Money

BFL monetizes FLUX through pay-as-you-go API use and separately licensed self-hosted deployments, with no API subscription or seat fees. API requests include commercial rights, while customers pay only for generated outputs.

Official API pricing varies by model and output size. FLUX.2 [klein] starts at $0.014 per image for 4B and $0.015 for 9B. FLUX.2 [pro] starts at $0.03 per generated image and $0.045 per edit. [max] starts at $0.07, and [flex] starts at $0.05. Earlier FLUX.1 Kontext [pro] costs $0.04 per image, while [max] costs $0.08.

BFL also sells self-hosted licenses: Builder is self-serve, while Platform, Professional, and Enterprise require sales contact. The public licensing page describes usage allowances but does not publish license prices. Enterprise API offers custom pricing, volume discounts, service-level agreements, and dedicated support.

BFL announced a $300 million Series B in December 2025 at a $3.25 billion post-money valuation. Sacra estimated $96 million in annualized revenue by August 2025. This is an analyst estimate, not a company-reported figure. No verifiable public total-addressable-market estimate or audited revenue disclosure surfaced in the reviewed sources.

Impact & Recommendations

Black Forest Labs raised $300 million at a $3.25 billion valuation in December 2025, a strong growth signal. FLUX has moved beyond an open model ecosystem into enterprise creative workflows. Black Forest Labs reports millions of users, a vendor-reported figure. Adobe offers FLUX in Firefly. Bloomberg reported Meta signed a $140 million multiyear contract. These placements signal strong distribution and enterprise interest.

Its advantage is a two-sided model: open weights support experimentation and self-hosting, while APIs and enterprise integrations support production use. FLUX.2 supports editing with up to ten reference images, useful for consistent visual assets. However, FLUX.2 [dev] has 32 billion parameters and a non-commercial license. Variants use different terms, so commercial buyers need to review licenses carefully. The large model also requires substantial graphics processing unit (GPU) memory for local use.

The closest alternatives occupy distinct lanes. Midjourney sells subscriptions for image generation. OpenAI offers image generation through an API. Stability AI’s Stable Diffusion 3.5 emphasizes customizable models that run on consumer hardware. Together, these rivals create pressure from established creator workflows, API distribution, and open alternatives.

Best fit: creative teams and developers that need image generation, editing, and API access. They can choose hosted or local deployment. Recommendation: shortlist FLUX for production visual workflows, then benchmark output quality and cost against competitors and confirm license terms. Outlook is positive, supported by funding and platform adoption. Long-term advantage depends on model quality, reliable integrations, and clear commercial licensing.

Products & Features

Teams can use FLUX through a browser Playground, a REST API, or downloadable weights for local deployment. The Playground supports image generation and editing, while the API uses asynchronous submit-and-poll requests with Python, TypeScript, and cURL guides.

FLUX.2 combines image generation and editing, supports up to 10 reference images, and produces edits up to 4 megapixels. Its hosted lineup includes [pro], [max], [flex], and [klein]; [dev] is local-only through BFL’s API pricing guide. FLUX.2 [klein] also runs locally, with 4B and 9B variants under different licenses. BFL now offers FLUX 3 Video through its dashboard and API. FLUX 3 Image is also available. FLUX.1 Kontext and FLUX1.1 [pro] Ultra remain listed API endpoints.

BFL uses pay-as-you-go pricing without subscriptions or seat fees. FLUX.2 [klein] starts at $0.014 per image. [pro] starts at $0.03 for generation and $0.045 for editing. [max] starts at $0.07, and [flex] starts at $0.05. FLUX 3 Image ranges from $0.041 at 768 pixels to $0.607 at 4K.

FLUX 3 Video starts at $0.17 per second for HD full renders. Draft mode starts at $0.06 per second. Older options include FLUX.1 Kontext [pro] at $0.04 and [max] at $0.08. FLUX1.1 [pro] Ultra costs $0.06 per image. Self-hosted commercial licenses and enterprise API plans are available through custom or tiered licensing. FLUX.2 [dev] is also available through FAL, Replicate, Runware, and Together AI endpoints.

User Feedback & Adoption

Black Forest Labs reports millions of FLUX users. Public feedback remains scattered across model forums. Its December 2025 announcement names Adobe, Canva, Meta, and Microsoft as partners. It also reports FLUX use on Fal.ai, Replicate, and Together AI. These are company-reported adoption claims, not independently audited user counts.

G2’s “Flux Pro AI” listing says it has no reviews, so it offers no dependable rating or review trend. The listing appears to cover a third-party product using FLUX, not Black Forest Labs itself. This research found no comparable FLUX rating on Capterra or TrustRadius.

Hands-on feedback highlights prompt adherence and text rendering as strengths. Tom’s Guide praised FLUX 1.1 Pro’s speed and prompt adherence. Another comparison favored Midjourney for human realism and skin texture. Results vary by prompt and comparison task, so these tests do not establish a universal winner.

Kontext user discussions show a more mixed experience. Individual Reddit commenters praise its editing potential. They also report inconsistent edits, resolution limits, artifacts, and framing problems. These posts are anecdotal, not representative survey results.

Local use requires compatible hardware and setup. Some users also report repeated retries when editing images. Black Forest Labs places Kontext Pro inside Photoshop. This can ease setup for Photoshop users. The announcement provides no independent satisfaction data.

Who It's Built For

Enterprise agreements are available from 200,000 image generations monthly, while FLUX also serves independent artists, developers, and creative teams. The audience spans solo creatives using open models, software teams embedding image generation, and companies building high-volume visual workflows. BFL offers managed APIs, self-hosted open weights, and custom enterprise deployments for teams needing private infrastructure or tailored models.

Creative professionals use FLUX for image creation and editing, campaign concepts, product mockups, and brand-consistent assets. FLUX.2 supports editing with up to 10 reference images and outputs up to 4 megapixels. Fine-tuning can adapt models to a brand or style from as few as 1 to 5 example images. Photoshop users can access FLUX.1 Kontext [pro] in Generative Fill. They can then refine images with Photoshop’s tools.

Developers use the API to add image workflows to products. They can run open-weight models through ComfyUI or Hugging Face Diffusers. Business teams can build production pipelines for image editing, outpainting, erasing, and virtual try-on. The strongest documented vertical fit is media and entertainment: BFL describes BurdaVerlag using customized FLUX models across its brands.

Fashion and luxury retail also appear in BFL’s enterprise examples. The company reports an unnamed luxury-brand deployment scaling seasonal output from 1,000 to 100,000-plus images. That result is vendor-reported.

The company serves both individual creators and large organizations, with no single employee-size cutoff stated. Its enterprise API volume threshold points to high-throughput teams, while open models and developer integrations support experimentation. BFL offers multi-region API availability, and Adobe made FLUX.1 Kontext [pro] available in Photoshop beta to users worldwide. The evidence indicates global reach rather than a single-country market focus.

Trust & Compliance Posture

Enterprise customers can choose a hosted API with zero data retention, dedicated endpoints, or self-hosting inside their own network. BFL’s enterprise page lists SOC 2 Type II, ISO 27001, and GDPR-compliant processing. It offers security documents and vendor assessments on request. It does not publish certificates or audit reports there.

BFL offers single sign-on (SSO) through SAML or OIDC providers, including Okta, Azure AD, and Google Workspace, with verified-domain enforcement. Custom deployments can use dedicated instances with zero retention and network isolation. BFL says self-hosting keeps inference inside the customer’s infrastructure, with no external calls.

BFL’s privacy policy says it collects account details, prompts, uploaded files, generated outputs, and device and usage data. For standard services, the policy permits use of de-identified data to improve services. Users can object by contacting BFL. Enterprise contract data falls outside the general privacy policy and is governed by customer agreements. The policy says information from the EEA, UK, and Switzerland may be transferred and processed elsewhere. The reviewed materials do not specify encryption methods, exact retention periods, or regional storage locations.

BFL says hosted services filter sexual content, hate, violence, gore, and self-harm, and add provenance metadata to outputs. Its policy prohibits CSAM and non-consensual intimate imagery, and bars tampering with C2PA credentials or watermarks. BFL reports two safety partnerships. It works with the Internet Watch Foundation to filter known CSAM from training data. Cinder evaluates FLUX.2 safety. Its responsible-development policy acknowledges that open-weight models can run outside BFL’s oversight, which limits the company’s control after release.

Tags & Categories

FLUX.2 generates images from text and edits existing images with multiple references. The playground supports up to 10 reference images, and BFL offers models through an API and as downloadable open weights.

Primary Category: AI Design Categories: AI Design Tags: text-to-image, image editing, image-to-image, generative AI, open weights, image generation API, developer tools, visual design, reference-image editing, marketing creative

Chiri Analysis

Chiri Score: 82/100

Verdict

Best for:

• Software teams embedding image generation and editing through a pay-per-use API

• Agencies and brand teams producing consistent assets with multi-reference editing and fine-tuning

• Enterprises generating 200,000 or more images monthly that need SLAs, SSO, and zero data retention

• Regulated or privacy-sensitive organizations that must self-host inference inside their own network

• Developers and researchers experimenting with open-weight models in ComfyUI or Hugging Face Diffusers

Not for:

• Hobbyists who want a flat-rate creator subscription instead of per-image billing

• Commercial users planning to deploy FLUX.2 [dev] without reviewing its non-commercial license

• Teams without high-memory GPUs that want to run the largest open models locally

• Buyers who require published security certificates or public self-hosted license pricing before contact

• Teams needing independent quality evidence for FLUX 3 Video before purchase

Head-to-head

Frequently Asked Questions

Is Black Forest Labs SOC 2 compliant?

Yes. BFL's enterprise page lists SOC 2 Type II, ISO 27001, and GDPR-compliant processing. Security documentation is available on request. BFL does not publish the certificates or audit reports on that page.

How much does FLUX cost?

The BFL API charges per generated image, with no subscription or seat fees. FLUX.2 [klein] starts at $0.014 per image. FLUX.2 [pro] costs $0.03 per generation and $0.045 per edit. FLUX.2 [flex] starts at $0.05, and [max] starts at $0.07. Enterprise and self-hosted license pricing requires sales contact, except the self-serve Builder tier.

Is Black Forest Labs good for enterprise?

Yes. Enterprise API agreements start at 200,000 generations per month. They include custom pricing, volume discounts, SLA guarantees, and dedicated support. SLA means service-level agreement, a contractual uptime and support commitment. BFL also offers SAML or OIDC single sign-on, dedicated instances, and network isolation.

Can I self-host FLUX models?

Yes. BFL releases open-weight models, meaning users can download the trained model files and run them locally. Self-hosting keeps data and inference inside the customer's infrastructure without external calls. Commercial self-hosted use requires a license tier: Builder, Platform, Professional, or Enterprise. Some open variants, such as FLUX.2 [dev], carry non-commercial terms.

Does FLUX API output include commercial rights?

Yes. Images generated through the BFL API include commercial rights. Self-hosted open-weight models follow separate license terms per variant. Buyers should review each model's license before commercial deployment.

Who are Black Forest Labs' competitors?

The main competitors are Midjourney, OpenAI's image generation API, and Stability AI's Stable Diffusion 3.5. Midjourney sells creator subscriptions. OpenAI distributes image generation through its API. Stability AI competes on open, customizable models that run on consumer hardware.

What can FLUX.2 do?

FLUX.2 generates images from text prompts and edits existing images. It accepts up to 10 reference images for consistent outputs. It produces outputs up to 4 megapixels. The FLUX Pro Finetuning API adapts models to a brand or style from 1 to 5 example images.

Who founded Black Forest Labs?

Investor Earlybird lists ten founders, including Robin Rombach, Andreas Blattmann, and Patrick Esser. These three helped create Stable Diffusion. Robin Rombach is co-founder and CEO. The company launched in August 2024 with a $31 million seed round led by Andreessen Horowitz.

---

Reviewed by Chiri Atlas Research Desk (AI Tooling Analyst) on 2026-10-03.