Text-to-Image Generation on MultiHuman-Testbench text-to-image benchmark leaderboard · SOTA2 Research

Text-to-Image Generation on MultiHuman-Testbench text-to-image

0.56DINO DiffSD3.5-Large

Evaluation Results

Method	Links
SD3.5-Large2026.05		0.56	0.63	0.41	78	86	39	88
FLUX [dev]2026.05		0.57	0.61	0.42	77	78	41	85
SDXL2026.05		0.62	0.62	0.49	78	77	37	83
RealCompo2026.05		0.62	0.64	0.52	57	50	15	77
CreatiLayout2026.05		0.63	0.64	0.48	75	78	38	84
FLUX Kontext [dev]2026.05		0.64	0.65	0.51	76	82	44	86
PeopleComposer2026.05		0.64	0.66	0.55	80	84	55	89
FLUX Fill [dev]2026.05		0.65	0.66	0.58	80	79	36	85