Images

Flux - review

Name: Flux - review
Item: Flux
Rating: 7.2
Author: Piotr Olszewski

7.2/10

Schnell is free and under the permissive Apache 2.0 license, dev has open weights for non-commercial use, and pro is available for a fee through API (billing per generated image) and from hosting partners. Locally you only pay with electricity and hardware. API prices and license terms can change - be sure to check the current pricing and license on the Black Forest Labs website before commercial deployment.

Last updated: Nov 6, 2026

Try Flux ↗

Rating: 7.2/10.

Flux is a family of image generators from Black Forest Labs-a team that comes from the people who previously stood behind Stable Diffusion. For me, it's today's strongest open-weights model that genuinely treads on the heels of Midjourney and DALL-E in terms of realism and correct text rendering in images. It comes in three variants: schnell (fast, most liberal license), dev (open weights for non-commercial use), and pro (best quality, API only). If you care about something you can run locally or integrate your own way, this is the first address I'd knock on.

Strengths

Photographic realism stands at a very high level-skin, textures, lighting, and depth of field look credible, and the model rarely falls into that plastic, over-smoothed look that gives away AI generation at first glance. This is today one of the stronger options for realistic portraits and scenes.
Text rendering in images is among the best on the open market-captions on signs, labels, or posters actually form legible words rather than gibberish of random letters, which was the bane of older diffusion models.
Open weights of the schnell and dev variants give real control: you run locally on your own GPU, build LoRAs, fine-tune to your style, and don't pay for every image or send data to someone else's server. For business and privacy, this is a huge advantage over closed competitors.
Excellent understanding of complex prompts - the model sticks to descriptions of multiple objects, their placement, and spatial relationships better than most open-source competitors, so you rarely have to fight with it over what should actually be in the frame.
Mature ecosystem - ComfyUI, Diffusers, ready-made workflows, tons of LoRAs and tutorials in the community. Jumping into Flux is much easier than into many niche models, because practically everything you need has already been described and integrated by someone.

Weaknesses

Licenses are varied and can bite - the dev variant is for non-commercial use only, and commercial use requires a separate license or reaching for pro/API. You need to read the terms carefully, because it's easy to fall into a trap using a model for profit that doesn't cover it.
Hardware requirements for the dev variant are substantial - it's a model with tens of billions of parameters, so for comfortable local work you want a GPU with a really large amount of VRAM. On a weaker card, either forget it or struggle with quantized versions and quality compromises.
Hand anatomy and fine details can still be unreliable - despite the excellent overall level, typical AI generation mishaps occur (fingers, jewelry, facial symmetry in the background), so for production you still need to select and sometimes retouch.

What forWorks best where you care about photorealism and correct text in images: product photos, mockups, posters, social media graphics with text, marketing materials. Shines strongly in applications where you need control and local execution - your own pipeline, fine-tuning for brand style, mass generation without paying per image. It's also a good foundation for building your own tool or image SaaS thanks to open weights.

Who is this forFor developers, studios, and creators who want real control over image generation - locally, with fine-tuning, without dependence on someone else's API. Less for the average user who just wants to type a prompt in a browser and get a nice image without configuration.

AlternativesIf you want a maximally artistic, polished look without configuration hassle - Midjourney. If strict adherence to a complex prompt and integration with the chat ecosystem matters - DALL-E in ChatGPT or Imagen from Google. If you need a fully open, commercially free alternative and aren't chasing absolute top quality - Stable Diffusion 3.5 or SDXL. Reach for Flux when you want to combine top open-source realism with local control and fine-tuning.

VerdictRating: 7.2/10. Flux is for me currently the best image generator with open weights and one of the few that genuinely competes with the closed-source top tier in terms of realism and text in images. It beats other open-source models hands down and gives something that neither Midjourney nor DALL-E can: full control, local execution, and fine-tuning. The downside is the entry barrier (hardware) and licensing nuances of the dev variant, which you need to figure out before commercial use. I recommend without hesitation to anyone building their own image tools or needing generation without dependence on someone else's API - for them it's the first choice today. For the average user who just wants a nice image from a browser without configuration, Midjourney will be easier.