Show HN: Agate, a 260M image model with separate thinker and renderer

Category: ai-ml

Tags: text-to-image, comfyui, generative-ai, diffusion-models, logo-generation

Score: 6.5/10 (Innovation: 6, Technical: 7, Documentation: 8, Utility: 5)

Agate is a 260M-parameter text-to-image model from LogoLabs, released with ComfyUI integration nodes that allow sampling in SD 1.x latent space. It stands out for its small size, built-in prompt pipeline, and AI-generated output marking, targeting logo and symbol generation.

Target audience: AI researchers, generative artists, ComfyUI users, and developers interested in small-scale image generation

Repository: https://huggingface.co/Logolabs/agate-preview-001 · Python · MIT · 3 stars

View on Hacker News