Ideogram 4 Launches as Top Open-Weight Image Model, Outperforming Models Up to 80B Parameters

Ideogram 4 launches as a 9.3B parameter open-weight text-to-image model that outperforms rivals up to 80B parameters, offering 2K resolution, multilingual text rendering, and spatial layout control — now freely available on GitHub and Hugging Face.

• Ideogram 4, a 9.3B parameter open-weight text-to-image foundation model trained entirely from scratch, is now publicly available with inference code and weights on GitHub and Hugging Face.
• The model introduces a structured JSON prompting interface with best-in-class multilingual text rendering, bounding-box spatial layout control, color palette conditioning, and native 2K resolution support, powered by a fully single-stream Diffusion Transformer architecture using Qwen3-VL-8B-Instruct as its text encoder.
• Benchmarks and third-party evaluations show Ideogram 4 ranks as the top open-weight image model across design, typography, and general-purpose leaderboards, outperforming much larger open models like FLUX.2 dev (32B) and HunyuanImage 3.0 (80B MoE) while competing closely with proprietary models from OpenAI and Google.