Flux is Black Forest Labs' family of image-generation models — the default open-weight choice for developers who want photorealistic quality without vendor lock-in. FLUX.1 [schnell] is free and commercially usable, [dev] is free for non-commercial use, and FLUX.2 [pro] is API-only with megapixel-based pricing (1 credit = $0.01). Best for developers self-hosting or building image generation into products.
Affiliate link — we may earn a commission
This analysis is based on documentation, public user reports, and vendor materials — not yet on our own hands-on testing. How we rate
Flux is Black Forest Labs' family of text-to-image models — built by the researchers who originally created Stable Diffusion, and now the default open-weight choice for developers who want photorealistic quality without lock-in. Free open variants (schnell, dev) plus a closed FLUX.2 API, priced from a penny a credit. Known for photorealism, strong text-in-image rendering, and character consistency. Best for developers building image generation into products and technical artists who self-host and fine-tune.
The key framing, like with any open-model tool: Flux isn't primarily a consumer app you click around in. It's a set of models. You either self-host the open variants, hit the API, or use a third-party platform that wraps it. That makes it a developer's tool first — the appeal is control, self-hosting, fine-tuning, and no vendor lock-in, not a polished UI.
What's changed for 2026 is FLUX.2. Released November 2025, it's a 32B-parameter flagship doing up to 4-megapixel generation and editing with accurate text and consistent characters. Pricing moved to a megapixel model — you pay by output resolution. If an older review only mentions FLUX.1, it predates this.
Flux fits developers integrating image generation into products, technical artists who want to self-host and fine-tune with LoRAs, teams needing character/style consistency across many images, and anyone who specifically needs reliable text rendering inside generated images. The open weights make it the go-to when control and no lock-in matter more than a nice interface.
It's also a fit for cost-conscious high-volume users — self-hosting schnell means zero marginal cost per image.
Where it's not the right pick: artists who want the best out-of-the-box creative tool with a polished UI (Midjourney), non-technical users (self-hosting and API work assume developer comfort), or anyone who wants a hosted product with support rather than open models plus community docs.
| Tool | Open weights | Self-host | Text rendering | Consumer UI | Cost |
|---|---|---|---|---|---|
| Flux | ✅ schnell/dev | ✅ Yes | ✅ Strong | ⚠️ Via third parties | Free / API from $0.01 |
| Midjourney | ❌ Closed | ❌ | ⚠️ Improving | ✅ Polished | $10+ |
| DALL-E 3 | ❌ Closed | ❌ | ✅ Strong | ✅ Via ChatGPT | $20 (Plus) |
| Stable Diffusion | ✅ Open | ✅ Yes | ⚠️ Weaker | ⚠️ Via tools | Free / compute |
| Ideogram | ❌ Closed | ❌ | ✅ Best | ✅ Yes | ~$20 |
Benchmarks and pricing checked August 2026. Competitor prices are approximate.
Flux vs Midjourney: Midjourney is a polished hosted product with a refined aesthetic and easy UI, but closed. Flux matches or beats it on photorealism and text rendering and is open-weight — self-host, fine-tune, no lock-in. For a ready creative tool, Midjourney; for building into products or self-hosting, Flux.
Flux vs Stable Diffusion: Both are open and self-hostable — and they share DNA, since BFL's founders created Stable Diffusion. Flux is generally the stronger current model, especially on photorealism and text. For the latest open quality, Flux; SD still has a vast ecosystem and older-hardware options.
Flux vs Ideogram: Ideogram is the specialist for text-in-image and comes as a polished app. Flux also renders text well and is open/self-hostable. For the absolute best text rendering in a ready app, Ideogram; for open weights plus strong text, Flux.
Flux pricing depends entirely on variant and access:
Checked August 2026 against BFL's docs. Same pricing for API and Playground. FLUX.2 cost scales with resolution — use BFL's calculator for exact figures. Third-party platforms that host Flux set their own prices.
The value story splits by who you are. For a developer who can self-host, schnell is genuinely free at any volume beyond your compute — unbeatable for high-throughput generation. For production apps needing top quality, the FLUX.2 API's per-penny credit pricing is competitive, though megapixel-based costs mean high-resolution work adds up. The honest trade, same as any open model: you get openness, control, and low cost, but you give up the polish and support of a hosted consumer product. For the target user — a developer or technical artist — that trade is usually the whole point.
Image generation in a product: A developer integrates the FLUX.2 API into an app, paying per generation with megapixel-based pricing scaled to output size.
Self-hosted high-volume generation: A team self-hosts FLUX.1 schnell to generate images at zero marginal cost, avoiding per-image API fees entirely.
Fine-tuned brand style: A technical artist trains LoRAs on Flux open weights to lock a consistent brand or character style across outputs.
Text-in-image graphics: A creator uses Flux specifically for reliable readable text inside generated images, where many models fail.
Prototyping before production: A developer prototypes with free schnell/dev, then moves to the FLUX.2 API for top-quality production output.
Flux is the default open-weight image model in 2026 for good reason. It matches or beats Midjourney and DALL-E 3 on photorealism and text rendering, and it's open — self-host, fine-tune, avoid lock-in, generate at zero marginal cost on the free variants. FLUX.2's November 2025 release (32B params, 4MP, consistent characters) keeps it at the frontier. For developers building image generation into products and technical artists who want control, it's the obvious choice.
The honest caveat is what open models always trade: Flux isn't a polished consumer product. There's no refined all-in-one UI from BFL — you self-host, use the API, or rely on third-party wrappers, backed by community docs rather than a help desk. For an artist who just wants the best ready-made creative tool, Midjourney is the better product. Flux is for people who build with models and value openness and control over convenience.
For developers, self-hosting, and no-lock-in image generation, strongly recommend. For a polished ready-to-use creative app, Midjourney fits better.
Note: Flux/Black Forest Labs open models have no consumer affiliate program. This rating reflects the models' genuine capability with no commercial incentive.
Best for: Developers building image generation into products, technical artists self-hosting and fine-tuning, teams needing character/style consistency, reliable text-in-image work Not ideal for: Artists wanting a polished ready-made tool (use Midjourney), non-technical users, anyone wanting a hosted product with official support Bottom line: The open-weight image model developers reach for first — photorealistic, strong text, self-hostable, now on FLUX.2. Excellent for building and self-hosting; Midjourney wins as an out-of-the-box creative product.
Partly — it depends on the variant. FLUX.1 [schnell] is fully open-source (Apache 2.0) and free for commercial use. FLUX.1 [dev] is open-weight but free only for non-commercial use. The higher-fidelity tiers — FLUX.2 [pro] and variants — are closed and API-only, priced per generation. So you can run Flux completely free if you self-host schnell; commercial production on the best models means paying API fees or licensing dev commercially.
Black Forest Labs uses credit-based API pricing where 1 credit = $0.01 USD, same price for API and Playground. Older FLUX.1 pro models run roughly $0.04-0.055 per image. FLUX.2 uses megapixel-based pricing — cost scales with output resolution, with the first megapixel at a flat rate and each additional MP adding to the total. Multi-image workflows with FLUX.2 Pro add ~$0.015 per input image. Use BFL's pricing calculator for exact costs at your resolution.
FLUX.2 is Black Forest Labs' latest flagship, released November 2025 — a 32-billion-parameter model that generates and edits images up to 4 megapixels with accurate text rendering, precise color matching, and consistent character identity across outputs. It uses a latent flow-matching architecture pairing a Mistral-3 vision-language model with a rectified flow transformer. There's also FLUX.2 Klein, a smaller/cheaper variant. FLUX.2 pricing is megapixel-based, so higher resolution costs more.
Match the variant to the job. FLUX.1 [schnell] — fast, fully open, commercial-OK — is best for high-volume or budget self-hosted generation. FLUX.1 [dev] gives higher open-weight quality for non-commercial work (or commercial with a license). FLUX.2 [pro] via API is the highest fidelity for commercial production. Developers prototyping or self-hosting start with schnell/dev; production apps needing top quality use the FLUX.2 API.
Different tools for different users. Flux matches or exceeds Midjourney on photorealism and text rendering in benchmarks, and it's open-weight — you can self-host, fine-tune, and avoid lock-in, which Midjourney can't offer. Midjourney is a polished hosted product with a refined aesthetic and easy UI. For developers building into products or self-hosting, Flux. For an artist who wants the best out-of-the-box creative tool, Midjourney.
Yes — that's a core appeal. The open-weight variants (schnell, dev) run on a capable GPU, letting you self-host, fine-tune with LoRAs, and generate at zero marginal cost. This is why Flux is the default for technical artists and developers who want control and no per-image fees. The closed FLUX.2 pro tiers are API-only, but the open variants cover self-hosting, and the community has built an extensive fine-tuning ecosystem around them.
The leading AI voice platform — text-to-speech, voice cloning, and conversational AI in 70+ languages.
The industry-standard collaborative design tool, now with AI features for UI generation, copy rewriting, and design automation.
AI image upscaler that adds realistic detail rather than enlarging pixels. The professional choice for photographers, designers, and AI artists.
