ACX complex compositionality
Comparison
5 prompts
·
42 models
Raters are asked: Select the image you prefer as a completion of the prompt: “{prompt}”
Completed
Elo scores
Each bar spans a 95% credible interval; the dot marks the mean. Hover for exact values.
Cost vs. Elo score
Generated images
“A 3d render of an astronaut in space holding a fox wearing lipstick”
GPT Image 2
Nano Banana 2 (Gemini 3.1 Flash)
Nano Banana 2 (Gemini 3.1 Flash)
riverflow-v2-pro
GPT Image 1.5 (High Fidelity)
Seedream 3.0
GPT-5 Image Mini
Flux.2 Pro
Gemini 2.5 Flash Image
Seedream 4.5
Seedream 5.0 Lite
GPT Image 1 Mini
Krea 2 Large
Nano Banana Pro (Gemini 3 Pro)
Kling V3
Qwen Image 2
Seedream 4.0
Gemini 3 Pro Image
Recraft V4.1
Recraft V4
Flux.2 Flex
Flux.2 Max
Flux.2 Dev Turbo
riverflow-v2-fast
Krea 2 Medium
Grok Imagine Image
Flux.2 Klein 9B
Nano Banana (Gemini 2.5 Flash)
Flux.1 Schnell
Flux.2 Klein 4B
Bria FIBO
GLM Image
MiniMax Image
Flux.1 Dev
Luma Photon
Flux 1.1 Pro
Qwen Image
Recraft V3
SD 3.5 Large
Wan T2I