Skip to content
AI ToolsOf The Trade

2026 Review

FLUX

Black Forest Labs' frontier image models — open-weight Klein to 4K FLUX.2 Max — priced per image.

ImagesFreemiumAPIPopular popularity (curated reach tier)Editorial score: 4.4 out of 585% positive — based on reviews across Reddit, Product Hunt

Facts checked Aug 6, 2026

About FLUX

FLUX is the image model family from Black Forest Labs, the German lab founded in 2024 by the original Stable Diffusion researchers. The FLUX.2 generation (November 2025) spans open-weight Klein models you can self-host to the hosted FLUX.2 Max flagship with 4K output and multi-reference conditioning across up to 10 images; the multimodal FLUX 3 (July 2026) adds 20-second video with audio in early access. There are no subscriptions — the API and browser Playground charge per image, from about $0.014 (Klein) to $0.07 (Max).

At a glance

Founded
2024
Headquarters
Freiburg, DE
Platforms
Web, Windows, macOS, Linux

Integrates with

ComfyUIHugging FaceOpenRouterfalReplicate

Best for

API image generationSelf-hosted modelsProduct integrationsCharacter-consistent art

Pros

  • The most widely adopted open-weight image models — the de facto successor to Stable Diffusion
  • No subscriptions or seats: pay only per generated image, from $0.014
  • Multi-reference inputs keep characters and brand styles consistent across generations
  • Well-funded ($450M+) and shipping fast: FLUX.2 Nov 2025, FLUX 3 Jul 2026

Cons

  • No flat-rate plan — high-volume costs are unpredictable next to Midjourney's unlimited Relax mode
  • Playground is a developer sandbox, not a polished consumer app
  • Top models (Pro, Flex, Max) are closed weights, hosted only

Our verdict

Best for: Developers and studios who want frontier image quality through an API, or free open weights to self-host.

Skip if: You want a consumer app with a subscription, community, and creative tooling — Midjourney or Freepik fit better.

Our review

Editorial score: 4.4 out of 5

Black Forest Labs is what happens when the researchers behind Stable Diffusion start over with funding and focus: FLUX.2 is arguably the best image model you can call through an API, and unquestionably the best you can download. Multi-reference conditioning — feeding up to 10 images to hold a character or brand style steady — solves the consistency problem that keeps generative images out of production pipelines, and per-image pricing with identical rates in the Playground and API makes costs legible in a way subscription tools never are.

What FLUX is not is a consumer product. There's no community feed, no mobile app, no unlimited plan; the Playground is a prototyping surface, and at serious volume, per-image billing can overtake Midjourney's flat $30 Standard tier. The flagship models are also closed weights, so the 'open' story really belongs to the Klein tier — excellent, but a step below Max.

Verdict: the first choice for developers integrating image generation and for anyone self-hosting in 2026. Creatives who want an app rather than an endpoint should look at Midjourney or Freepik — which, tellingly, serves FLUX models under its own subscription.

What reviewers elsewhere say

85% positive Developers praise output quality, prompt adherence, and the open weights; complaints focus on hosted-only flagship models and per-image costs at scale.

Sentiment estimated from reviews on Reddit, Product Hunt, as of Aug 2026. Third-party scores are point-in-time snapshots. We don't host user reviews.

Key features

  • FLUX.2 Max flagship: up to 4K generation with strong prompt adherence
  • Multi-reference conditioning — up to 10 input images for consistent characters and styles
  • Open-weight FLUX.2 Klein models for self-hosting and fine-tuning
  • Pay-per-image API and browser Playground (playground.bfl.ai) at identical prices
  • FLUX 3 multimodal model: images plus 20-second video with native audio (early access)
  • Powers image features inside Adobe, Canva, Meta, and Freepik products

How to use FLUX

  1. Create an account at bfl.ai and open the Playground, or grab an API key.
  2. Pick a model: Klein for cheap volume, Pro for everyday quality, Max for flagship output.
  3. Add up to 10 reference images to lock characters, products, or styles.
  4. Iterate in the Playground, then call the same endpoint from your code at the same price.
  5. For self-hosting, download Klein open weights and run them in ComfyUI or your own stack.

Pricing

Open weights$0

FLUX.2 Klein open weights free to download; licensing tiers for commercial self-host deployments.

API / PlaygroundPay-per-image

From $0.014/image (Klein) to $0.03 (Pro), $0.05 (Flex), and $0.07 (Max); credits at $0.01 each, cost scales with output megapixels.

Pricing verified August 2026 · prices as listed by the vendor; check the official site for current offers.

Try FLUX (opens in a new tab)

Frequently asked questions

What is FLUX?

FLUX is the family of image generation models from Black Forest Labs, founded by the original Stable Diffusion researchers. It spans open-weight models you can self-host and hosted frontier models (FLUX.2 Pro, Flex, Max) available per-image through an API and browser Playground.

How much does FLUX cost?

There is no subscription. API and Playground pricing is per image: roughly $0.014–0.015 for Klein models, $0.03 for FLUX.2 Pro, $0.05 for Flex, and $0.07 for Max, with cost scaling by output resolution. Open-weight Klein models are free to download.

Is FLUX open source?

Partially. The Klein models ship open weights you can run and fine-tune locally, while the flagship Pro, Flex, and Max models are closed and hosted-only. BFL has committed to releasing open variants of each generation.

FLUX vs Stable Diffusion — which should I use?

FLUX generates better images and follows prompts more faithfully; it's the newer work of the same researchers. Stable Diffusion still has the larger ecosystem of community checkpoints and tooling. New projects mostly start on FLUX.

Tags

imagesopen-sourcefine-tuningapi

Similar tools

Share