pants2
8 hours ago
No modern or over-time comparisons?
In some ways I prefer looking at these over the hyper-polished 2026 image generators tbh.
vunderba
7 hours ago
I've been running a benchmark for comparing generative image models with a strong emphasis on complex prompt adherence since the days of SDXL back in 2023.
It’s not really intended to be viewed chronologically but there is a progress metric that shows how models have improved their scores over time, which might help you compare earlier models with later models such as Ideogram4 or Moss-Image released in 2026.