Google DeepMind 負責人表示,Gemini 4 已進入後訓練初期,期望「遠早於」2026 年底推出初期版本。[13][19]
Google 打算在初期版本推出後快速改進;這不等於已宣布上市,也不能證明模型的推理、安全性或可靠度已達到特定水準。[13][19]
What did new Google DeepMind chief Koray Kavukcuoglu reveal at The Information’s AI Agenda Live Summit about Gemini 4’s early post-trainingAI-generated editorial illustration; not an image of Gemini 4 or the AI Agenda Live Summit.
AI 提示詞
Create a landscape editorial hero image for this Studio Global article: What did new Google DeepMind chief Koray Kavukcuoglu reveal at The Information’s AI Agenda Live Summit about Gemini 4’s early post-training. Article summary: Kavukcuoglu’s central message was that Gemini 4 had entered early post-training and that Google hoped to release an initial version “much earlier” than the end of 2026, then iterate quickly—not that it had announced a fi. Topic tags: general, general web, user generated, documentation. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks,
openai.com
Google DeepMind 日常營運負責人 Koray Kavukcuoglu 在 2026 年 9 月 23 日《The Information》舉辦的 AI Agenda Live Summit 上表示,Gemini 4 已進入後訓練初期。他希望「遠早於」年底推出初期版本,之後快速迭代。這是研發進度與推出目標,不是上市公告;Google 也沒有給出確切日期。1913