Qwen-Image-3.0 Released: Native Multilingual Support and Dense Layouts
aziz4ai · x · 2026-08-05
The image generation model Qwen-Image-3.0 is now available on the Runware platform, bringing significant improvements in text-image mixing and detail rendering.
The new model supports up to 4.5k tokens input, capable of generating dense multi-panel layouts in one pass. Its text rendering is highly enhanced, keeping text legible down to 10px and accurately generating LaTeX and mathematical formulas. Additionally, it natively supports 12 languages and achieves near-photographic detail for skin, pores, and hair.
More from Multimodal
- A New Knimare on Elmet Street - made with Wan2GP — ajrss2009 · 2026-08-05
- Image-to-Video Dominates? Developer Questions the Practical Value of Pure Text-to-Video — cocktailpeanut · 2026-08-05
- Running MiniMax Video Model Locally: M3 Ultra Takes 2.6 Hours for 10s Video — cocktailpeanut · 2026-08-05
- MiniMax Video Generation: Optimizing Fluid Dynamics via Detailed Prompts — Extreme_Speed6654 · 2026-08-05
- Fixing Quality Loss in MiniMax Image-to-Video: 16:9 Aspect Ratio Preserves Face Details — WearNatural5992 · 2026-08-05
- MiniMax H3 + LTX 2.3 Upscale Comparison: High-Quality Video Generation on Home GPU — Landrews-89 · 2026-08-05