

Ideogram 4.0
#20 en Texte-vers-imageIdeogram · v4.0 · depuis 2026-06-03 · 27× · vu le 31 juil. 2026
Ideogram 4.0 is Ideogram AI's first open-weight text-to-image model, released on June 3, 2026. It is built on a 9.3-billion-parameter single-stream Diffusion Transformer architecture (34 layers, trained from scratch) with Qwen3-VL-8B-Instruct as the text encoder. Key features include native 2K resolution (2048×2048 px), structured JSON prompting with bounding-box layout control, up to 16-color hex palette conditioning, and multilingual text rendering. Model weights are available on Hugging Face under the Ideogram Non-Commercial Model Agreement; inference code is Apache 2.0 on GitHub. Fine-tuning of the weights is permitted (free for non-commercial use; commercial use requires a separate paid license).
Fonctionnalités
| Max Resolution | Native 2048×2048 px (2K); supported resolutions: 256 to 2048 px per side (multiples of 16), aspect ratios up to 6:1 |
| Platform | Web app (ideogram.ai), Ideogram Developer API, Hugging Face (open weights: nf4 and fp8 quantizations), GitHub (inference code), ComfyUI (day-0 support), fal.ai (hosted endpoint) |
| Release Date | June 3, 2026 |