Logolabs Unveils Agate, a 260M-Parameter Text-to-Image Model Rivaling SD1.5

Logolabs has released Agate-003-preview, a sub-300M-parameter text-to-image model trained for around €5,000 and 145 GPU hours. Using a split "thinker" Transformer plus U-Net architecture, it claims SoTA results at its scale, rivaling SD1.5 with a third of the parameters.

2026-09-29 ~ 2026-09-29 · 2 related posts

• Logolabs' Agate: 260M-Param Text-to-Image Model Trained in Just 145 GPU-Hours, MIT Licensed — linoy_tsaban · 2026-09-29
• 300M-param image model trained for €5000 claims SD 1.5-level performance — incorporo · 2026-09-29