MiniMax releases open-weights H3-Image: 33.1B DiT generating 4K images in under 10 seconds
NielsRogge · x · 2026-11-11
MiniMax released H3-Image, an open-weight image generation model that tops Papers with Code trending:
- Architecture: a 33.1B diffusion transformer supporting text-to-image, reference-based generation (up to 16 reference images), and image editing
- Speed: generates 4K (4096×4096) images in under 10 seconds
- Quality: rich realistic details and strong prompt adherence
- Post-training pipeline: gains come from improvements across data, reinforcement learning, sampling-step reduction, prompt expansion, and deployment optimization — "no single change made the difference"
Code and weights are on Hugging Face (MiniMaxAI/MiniMax-H3-Image) with a full technical report.
Related event: MiniMax Open-Sources H3-Image: Native 4K Generation Aimed at Game Assets(74 posts)→
More from Multimodal
- Gemini-generated endless loop of quantum cryostats self-assembling — hapticdata · 2026-11-11
- Dad brings his daughter's crayon drawings to life with AI for $1.75, and it's unsettling — Illustrious_Ad_4292 · 2026-11-11
- MOSS-Image-SFT vs MiniMax-H3-Image (Pruned): 1:1 image benchmark across 10 art mediums — SkirtSwimming8950 · 2026-11-11
- UniPath (NeurIPS 2026): adaptively coordinating understanding and generation in one model — jindong_wang92 · 2026-11-11
- An AI-built RPG where a PhD student battles peer reviewers — s_scardapane · 2026-11-11
- Open-source node turns any ComfyUI workflow into a one-click model, 19 models bundled — jefoce · 2026-11-11