MiniMax 开源 H3-Image：33.1B DiT 十秒内生成 4K 图像，支持 16 张参考图
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.
所属事件:MiniMax 开源 H3-Image：原生直出 4K 图像(75 条相关)→
「多模态」频道最新
- Seedance 2.5 支持直接生成 4K 视频,免额外超分步骤 — charis_ai · 2026-11-12
- codec.audio 上线:多语言噪声环境下的神经音频编解码器横评 — kastnerkyle · 2026-11-11
- Linus Ekenstam 转发 AI 视频新作:这就是当下生成式视频的巅峰 — LinusEkenstam · 2026-11-11
- AI 视频现状盘点:生成式视频真实感已到新高度 — LinusEkenstam · 2026-11-11
- 用 MOSS-Image-Base 复刻 2000 年代闪光灯相机质感，11 条完整提示词公开 — SquareBat780 · 2026-11-11
- Premiere Pro 公测 Generative Remove：圈选即可从视频里抹除物体 — jnack · 2026-11-11