

Flux
FeaturedNot yet source-checkedLink check failingBlack Forest Labs' open-weights image model — rivals Midjourney quality.
Machine-drafted from the vendor's site · not yet source-checkedProfile updated Link checked How we rate
In short
Flux provides frontier-tier image quality with open weights for self-hosting and API access for production. It is best for developers needing high-quality, customizable image generation without relying solely on closed-source platforms.
Pick Flux when you want frontier-tier image quality with open weights or self-hosting options.
Skip it if you want a turnkey web app — Midjourney's UX is more polished.
Flux comes from Black Forest Labs, founded by the original Stable Diffusion team. Flux.1 [pro] competes with Midjourney v6/v7 on aesthetics and prompt adherence while [schnell] and [dev] are open-weights and self-hostable. The release reset the open-source image-gen landscape almost overnight.
For self-hosted production, Flux is the new default — better than SDXL on every axis, and trainable. For API access, Flux Pro is available via Replicate, Together, fal, and Black Forest Labs' own API at competitive per-image rates. Inside ComfyUI and the rest of the open-source tooling ecosystem, Flux support is now first-class.
The trade-offs: Flux Pro is API-only (no open weights for the top model), and self-hosting [dev] needs a serious GPU (24GB+ VRAM for comfortable inference). The community ecosystem of LoRAs and fine-tunes is younger than SD's but growing fast.
Flux is the most important image-gen release of the last 18 months. It reset the open-source ceiling, gave Stability a serious competitor, and made "self-hosted production image gen" a realistic choice again.
— The AI Tool Bible editorial team
Pros
- ✅ Open weights for [schnell]/[dev] variants
- ✅ Quality rivals Midjourney
- ✅ Excellent prompt adherence
- ✅ Self-hostable
Cons
- ⚠️ [pro] is API-only
- ⚠️ Self-hosting needs serious GPU
Facts
- Platforms
- api
- GitHub license
- Apache-2.0
- GitHub stars
- 26,059 (checked Nov 13, 2026)
- Last GitHub push
- Jul 31, 2025
- First commit
- Aug 2024
- Model used
- Flux.1 [schnell / dev / pro]
- Hacker News
- 8 stories · top 44 points · last Jan 16, 2025 · checked Nov 11, 2026
- Earliest Wayback capture
- 2024-08-01 (earliest archived capture of the site, not a launch date)
- Hugging Face
- 42 models · 5,461,139 downloads (30 d) · 44,356 likes for author black-forest-labs · checked Nov 11, 2026
Where this tool sits
Use cases
Frequently asked
- Who created the Flux image model?
- Flux is developed by Black Forest Labs, a company founded by the original Stable Diffusion team.
- Can I self-host the Flux model?
- Yes, the [schnell] and [dev] variants are open-weights and self-hostable, though [dev] requires a GPU with at least 24GB of VRAM for comfortable inference.
- How does Flux compare to Midjourney?
- Flux.1 [pro] competes with Midjourney v6/v7 on aesthetics and prompt adherence, while offering open-weights options that Midjourney does not.
- Is the top-tier Flux model available as open source?
- No, Flux Pro is API-only and does not have open weights, whereas the [schnell] and [dev] variants are open-weights.