whatmodel.sh / image / LLaDA-Image-Turbo

LLaDA-Image-Turbo

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

LLaDA-Image-Turbo is an open-weights image model from Ant Group, released September 2026. Its published size is 6B parameters. Licence: apache-2.0.

Model information sources

Release date ↗ · Weights licence ↗ · Parameter count ↗

6B is stated for the family, covering both Base and Turbo; the card does not break down which components (e.g. VQ tokenizer/decoder) are inside the 6B.

53.5±2.6#70 of 144 image modelsfrom 6 benchmark results across 2 sources

Score within the LMArena text-to-image comparison set only.

$ how it compares

The image models ranked either side of LLaDA-Image-Turbo, out of 197 with a score, then more from Ant Group.

#modelscorewhereRAMreleased
85Qwen-Image Alibaba54±1api $0.025/image +local~13GB2025-08
86HiDream-O1-Image-Dev HiDream54±5comparison only2026-05
87LLaDA-Image-Turbo Ant Group54±3local~22GB2026-09
88LongCat-Image Meituan53±4api $0.13/image +local~15GB2025-12
89Step1X-Edit v1.2 StepFun53±2local~51GB2025-11
more from Ant Group
79LLaDA-Image Ant Group55±3local~22GB2026-09

see all image models · find the ones that fit your machine

$ what it's good at

text to image▂▅▅▃▄▄█▃▁▂▁▁ #60 of 126 · 97% of its score · 6 tests

Beats FLUX.2 [klein] 9B, Qwen-Image here, though they rank higher overall. Trails LongCat-Image here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
UltraText Bench (Composite, bilingual dense text rendering) lab41.3% ✎▁▂▂█▃▁▂▃·▁·▁ beats 9/22hard · avg 28%▇▇▇▇▇▇▇▇▇▇ 26%sharp · at its level · noisy
LongText-Bench (EN) lab89.9% ✎▃·····▅██▅▃▃ beats 1/12very easy · avg 88%▇▇▇▇▇▇▇▇ 19%small run
LongText-Bench (CN) lab91.9% ✎▃·······████ beats 5/12very easy · avg 89%▇▇▇▇▇▇ 15%small run
GenEval lab82.0% ✎▇▃▅▇▁▄▅▃█▃▃▅ beats 31/39easy · avg 79%▇▇▇▇▇▇ 14%small run
CVTG-2K word accuracy lab84.0% ✎▄▄▄█▄▄·▄···▄ beats 0/8easy · avg 81%▇▇▇▇▇ 12%small run
DPG-Bench lab88.5% ✎▄·▂·▂▆█▄▂▄▃▂ beats 16/17very easy · avg 87%▇▇▇▇ 10%small run

$ where to run it

Other ways to run it

version

  1. Run it as published · needs ~21.6 GB · from github.com/leejet
    sd-cli --diffusion-model models/llada-image-turbo-q8_0.gguf --llm models/llada-image-turbo-text_encoder-q4_K.gguf --tokenizer models/tokenizer.json --vae models/llada_vae.safetensors --embeddings-connectors models/llada-image-turbo-connectors.safetensors --prompt "a lovely cat holding a sign says 'llada.cpp'" --width 1024 --height 1024 --steps 4 --cfg-scale 1.0 --seed 42 --output output.png

Weights and documentation: huggingface.co/inclusionAI ↗

$ how the score was worked out · LLaDA-Image: Building Strong… 71%, UltraText Bench: A Comprehens… 26%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipes · lab-reported5▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 71%CC-BY-SA-4.0
UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation (arXiv:2610.09823v1) · lab-reported1▇▇▇▇▇▇▇▇ 26%CC-BY-4.0
prior · pull toward the middle▇ 3%
weighted average= 53.5

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json