MiniMax Open-Sources H3-Image: 16MP High-Res Images in 10 Seconds at 50%-70% of Rivals' Price
赛博禅心 · wechat · 2026-11-11
MiniMax has released and open-sourced the weights of its image generation model MiniMax H3-Image, supporting text-to-image, image reference, and image editing:
- High resolution + speed: generates 4096×4096 (16MP) images within 10 seconds, roughly 5× faster than mainstream offerings (1.5-2× in standard mode, 3-4× in Highspeed mode).
- Realism: focused on eliminating the "AI look," with notably detailed and realistic output.
- Instruction following: strongest instruction-following among open-source models, supporting up to 16 reference images.
- Post-training investment: strengthened the entire post-training pipeline across data, reinforcement learning, step compression, and deployment optimization.
- Pricing: 50%-70% of mainstream tools in the same tier; 50% off for 16MP-and-above requests during the first week after launch.
Weights are available on Hugging Face, with API services offered via the MiniMax open platform, MiniMaxCode, and Hailuo.
More from Multimodal
- BaoCut author's 3-level confirmation workflow: preview HTML video before rendering — dotey · 2026-11-11
- Microsoft ships MAI-Image-2.6 Flash: 2.8x faster than GPT-Image-2-Medium, now Bing's default — JordiRib1 · 2026-11-11
- Animate real places in 3D with Claude and free map data from AWS and Overture — drbronk · 2026-11-11
- Fable 5.2 builds its own engine: 6cm FLIP wave sim and rigged galloping horses — AymericRoucher · 2026-11-11
- Redditors reproduce the disputed Fable 5.2 demo video, cost ~$20 in API tokens — timeport-0 · 2026-11-11
- Dev builds an AI paintbrush that repaints her first drawing in 55,185 strokes — repligate · 2026-11-11