LOADING DATASHEET
LOADING DATASHEET
by Black Forest Labs
RANKED #3 OF 90 IMAGES · OVERALL #13 OF 6,530 · VIBE SCORE 7.3 · 13,196 VOICES
An advanced image generator for creators wanting detailed, custom pictures.
1.6k mentions
920 mentions
26 weeks · 10,995 voices
HOLDING UP
Recent vibe is steady. No clear fade in the crowd.
Open weights. You can use a hosted service, or download it and run it yourself, free.
Open weights: download Flux once and it's yours. No subscription, no rate limits, works offline.
12B parameters, an FP8 file is about 12 GB, while a Q4 GGUF is about 6.5 GB
NVIDIA GeForce RTX 3060 12GB
12 GB VRAM
~$250 USEDQ4 GGUF or FP8 (low VRAM mode), ~15-25s per image (Schnell, 4 steps)
NVIDIA GeForce RTX 4070 Ti Super 16GB
16 GB VRAM
~$750 NEW / $650 USEDFP8 or Q8 GGUF with LoRAs, ~10-15s per image (Dev, 20 steps)
NVIDIA GeForce RTX 4090 24GB
24 GB VRAM
~$2,000 USED / $2,500Full FP16 precision, ~12-18s per image (Dev, 20 steps) with ControlNet
The three cards above are researched picks. Search the machine you actually have — we only say yes if it should reply at a usable speed, not just load the weights and crawl.
HARDWARE PICKS & PRICES ARE RESEARCHED FROM THE LIVE WEB AND REFRESHED AUTOMATICALLY. TREAT THEM AS BALLPARK, NOT GOSPEL.
Everyday jobs on each of those rigs: how long Flux takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMNVIDIA GeForce RTX 3060 12GB | SWEET SPOTNVIDIA GeForce RTX 4070 Ti Super 16GB | FULL POWERNVIDIA GeForce RTX 4090 24GB |
|---|---|---|---|
| Make an imageone picture at normal settings | ~20 S★★★★★ | ~13 S★★★★★ | ~15 S★★★★★ |
STARS ARE THE COMMUNITY SCORE FOR THAT KIND OF WORK, DOCKED FOR HOW SQUASHED THE WEIGHTS GET AT EACH TIER. TIMES ARE COMPUTED FROM RESEARCHED SPEEDS. BALLPARK, NOT GOSPEL.
Community members love Flux for its incredible prompt accuracy and detailed results, whether making art or everyday phone-style photos. However, some users find it slow, pricey on cloud services, or frustrating when hit with refusal filters and errors.
11169 POSTS · 2027 COMMENTS · REDDIT · HACKER NEWS · BLUESKY · DEV FORUMS · GITHUB · X · DEV.TO · LEMMY · BLOG · OTHER FORUMS · STACK OVERFLOW
50 thumbs up · 11 thumbs down
5 thumbs up · 47 thumbs down
1 thumbs up · 29 thumbs down
15 thumbs up · 13 thumbs down
5 thumbs up · 5 thumbs down
6 thumbs up · 0 thumbs down
“Flux Can Generate Really Good Fake, Low Quality And Boring Phone Photos!.”
“I am continuing FLUX LoRA training with 256 images experiments, the model learnt even my broken teeth :).”
“It's not merely "uncensored but lacking knowledge." It's the thing many people have been waiting for.”
“14 Mind Blowing examples I made locally for free on my PC with FLUX Kontext Dev while recording the SwarmUI how to use tutorial video - This model is better than even OpenAI ChatGPT image editing - just prompt: no-mask, no-ControlNet.”
“Kohya SS GUI very easy FLUX LoRA trainings full grid comparisons - 10 GB Config worked perfect - just slower - Full explanation and info in the comment - seek my comment :) - 50 epoch (750 steps) vs 100 epoch (1500 steps) vs 150 epoch (2250 steps).”
“Prompt adherence with Flux is insane.”
Checked picks first: tags say why you would switch, stars say how fully each one stands in for Flux.
Movement is almost human with KlingAi
Asked Flux Kontext to create a back view of this scene
why this flux so mad
Messing around with FLUX Depth
just curious what tools might be used to achieve this? i m using sd and flux for about a year but never tried video only worked with images till now
Veo3 + Flux + Hunyuan3D + Wan with VAce
You can train handwriting loras with flux
Huge FLUX news just dropped. This is just big. Inpainting and outpainting better than paid Adobe Photoshop with FLUX DEV. By FLUX team published Canny and Depth ControlNet a likes and Image Variation and Concept transfer like style transfer or 0-shot face…
Mirrorscapes - FLUX
I trained a handwriting flux fine tune
Made a dog smile with no code, now I'm looking for a friend like that. Anyone seen one?
Announcing Flux: The Next Leap in Text-to-Image Models