MiniMax 开源 H3-Image 生图模型，10 秒就能生成一张 4K 图像Chinese only
MiniMax发布图像生成模型MiniMax H3-Image并开放权重。
2 reports4 sources
Reports 2
量子位 AI
Selected
Machine translation
MiniMax releases H3-Image as an open-source image model with 16MP output
MiniMax has released H3-Image as an open-source image model supporting output up to 16MP (4096×4096 pixels) and input of up to 16 reference images. A QbitAI writer reports generating orthographic views from a product image in a hands-on test; the article also says API access is offered, with 16MP requests half-price in the first week.
ginobefun (@hongming731)
Same story
Machine translation
Gemini 4 Carbon expands access to paid developers and Ultra users
Gemini 4 Carbon is now available to more paid developers and Ultra users for complex coding, scientific research, and multi-agent tasks. Official materials also cover how it differs from Argon, task costs, and controlled safety access, giving teams more to assess before migrating.
7 more sources reported this
- IT之家 科技新闻
- Rohan Paul (@rohanpaul_ai)
- The Verge AI
- Latent Space AI
- 量子位 AI
- ginobefun (@hongming731)
- MarkTechPost AI 研究报道
SourceIT之家 科技新闻 ↗ginobefun (@hongming731) ↗IT之家 科技新闻 ↗量子位 AI 报道 ↗