Agate-001-preview: 260M-param open-weights image model nears SD 1.5 with thinker-renderer architecture
incorporo · reddit · 2026-09-26
Logolabs released Agate-001-preview, a 260M-parameter (including text encoder) image generation model under MIT license, claiming Pareto-frontier results that approach SD 1.5—a model with roughly 3x the total parameters.
Architecture highlights:
- A hybrid "thinker-renderer" design: a small recurrent transformer "thinker" reads the prompt and plans a 16×16 region map via cross-attention to a text encoder (based on Ettin-68M, frozen for 10 epochs then jointly trained), while an FDCM convolutional renderer converts latent thinking tokens into images.
- Trained 26 epochs on Flux-Reason-6M and not yet converged; currently limited to 256×256 with the SD-1.5 VAE.
Use cases: cheap synthetic data generation (adaptable to generative denoising/editing), training small models on proprietary datasets, thumbnails/search loaders, and on-mobile generation.
Weights and a WebGPU demo are live on Hugging Face.
Related event: 260M-Parameter Open Model Agate-001 Nears SD 1.5 Performance(2 posts)→
More from Multimodal
- Sophon-Realtime-v1, a full-duplex speech-to-speech model, trends on Hugging Face — adventists-ai · 2026-11-11
- Asked Fable 5.2 to 'go wild' making a game, got a Pong clone — and loved it — ariG23498 · 2026-11-11
- Western town built with Fable 5.2: ray-traced weathering ages every board — ravisparikh · 2026-11-11
- Diffusion Week 2026 releases all 10 lectures on video generation, world models and more — LucaAmb · 2026-11-11
- Running Wan 2.2 on a RTX 3080 12GB: 5B Breaks Down, 14B Works Surprisingly Well — fixxxer2606 · 2026-11-11
- Indie dev teases Z-Image-based model 'on par with latest closed-source models' — darlens13 · 2026-11-11