FLUX is the image model that most other image models quietly use as a benchmark. The 2024 release from Black Forest Labs broke the open-weights field, and the 2025 Kontext release added editing on top. In 2026, FLUX still sits at the top of the pure-image-quality bracket, even as Nano Banana Pro takes the editing-first crown and Midjourney v7 keeps the stylisation crown.
I spent a week running FLUX.1 [pro], [dev], [schnell], and Kontext across the kinds of jobs we ship at Vuela: product photography, brand campaigns, hero illustrations. Below: what each tier is for, where FLUX still leads, and where it has been overtaken since launch.
What is FLUX (and what is Kontext)?
FLUX is the text-to-image family from Black Forest Labs. The original FLUX.1 release in August 2024 shipped in three tiers: FLUX.1 [pro] (closed weights, top quality, API only), FLUX.1 [dev] (open weights, near-pro quality, non-commercial), and FLUX.1 [schnell] (open weights, distilled for speed, Apache 2.0).
FLUX.1 Kontext, released May 2025, added in-context image generation and editing: prompt with text and images together, and the model edits inside that context. It is the FLUX answer to the conversational editing trend Nano Banana started.
Access is wide: bfl.ai for the official playground, Hugging Face for weights, and almost every aggregator and inference platform exposes the model.
How I got access
I used the BFL Playground for [pro] generations, Hugging Face inference endpoints for [dev], and a local 4090 for [schnell] tests. Kontext I ran through the official API. Costs ranged from a few cents per image on aggregators to zero on local hardware.
The three jobs I tested
Same prompts across all FLUX tiers so I could compare the trade-offs.
- Hero product shot. A perfume bottle on a marble surface with motivated window light. Tier comparison: pro vs dev vs schnell.
- Editorial portrait. A character with a complex outfit, shallow depth of field, magazine style.
- Kontext edit chain. A starting still + three follow-up edits using FLUX.1 Kontext.
The test results
Test 1. Hero product shot
Prompt: “A glass perfume bottle on a polished Carrara marble surface, soft directional window light from camera left, shallow depth of field. Editorial campaign style. 8K.”
On [pro], the bottle had crisp specular highlights, accurate glass refraction, and a marble surface that read as real Carrara rather than fake granite. On [dev], the result was about 90% there with slightly less reflection detail. On [schnell], speed was the headline (under two seconds per render) at noticeably softer texture. For client work [pro] is still the answer; for fast iteration [schnell] is genuinely usable.
Test 2. Editorial portrait
Prompt: “A young woman in a complex layered outfit, freckled face, freshly cut blunt bob, looking off-camera, magazine editorial lighting. Shot on medium format. 4K.”
Skin texture and hair detail are where FLUX has always quietly led. [pro] held freckle placement, hair direction, and fabric layering at a level Midjourney v7 matches stylistically but exceeds in painterly territory. For realistic editorial, FLUX still wins.
Test 3. Kontext edit chain
Prompt: “Start: a sneaker on a white background. Then: "place the sneaker on a beach at sunset." Then: "change the laces to red." Then: "add motion blur as if it is being thrown."”
Kontext handled the chain cleanly. The sneaker identity (silhouette, brand cues) survived all three edits. The lace colour change stayed local — the rest of the image did not shift. Compared to Nano Banana Pro on the same chain, Kontext is slightly stronger on photographic detail and slightly weaker on conversational text-based edits. The two are complementary, not competing.
The annoying parts
[pro] is closed. The best tier is API-only. For teams that want to self-host the top model, FLUX is not the answer.
[dev] is non-commercial. The strongest open tier has a non-commercial licence. Commercial use needs a separate licence or the [pro] API.
No pipeline. FLUX is a model, not a platform. Cloning, translation, motion, and video pipelines all need other tools on top.
Is it worth the price?
For agencies needing the best photographic image quality, FLUX.1 [pro] is still the model to reach for first. For high-volume social or batch work, [schnell] at near-zero per-image cost is the right call.
For developers integrating into a product, per-image pricing across aggregators is in the cents range — predictable and easy to budget.
How Vuela.ai fits into a FLUX workflow
FLUX is the image-quality backbone of many production pipelines, including the one Vuela.ai exposes to creators. Where FLUX ends, Vuela picks up: turning a FLUX-generated still into a video ad, cloning a viral format around it, translating the finished asset across languages with real lip sync.
Vuela.ai bundles FLUX-class image quality with video, voice, cloner, and translator under one flat plan. No need to juggle BFL credits, Hugging Face quotas, and a separate video vendor.
FLUX-quality images inside a real content pipeline
Vuela.ai gives you FLUX-grade image quality plus video, voice, cloner, and translator on one flat plan.
The verdict
FLUX is, in May 2026, still the strongest text-to-image foundation in the market. For pure photographic generation, [pro] leads. For editing-first workflows, Nano Banana Pro now has a competitive answer with Kontext close behind. For stylised aesthetics, Midjourney v7 keeps the crown.
In a 2026 stack the right play is to use FLUX where photographic fidelity matters and stitch it into a platform that handles the rest of the pipeline. That platform is Vuela.ai.