Workflow
Gemini 3 Flash
icon
搜索文档
Google unveils 'Gemini 3 Flash' AI model focused on speed and cost
Youtube· 2025-12-18 00:42
We do have some breaking news on Google and for that we're gonna go to dear DOS. Hey De. >> Hey good morning Carl.So Google just announcing its latest AI model called Gemini 3 Flash. Quote frontier intelligence built for speed at a fraction of the cost. Now it prioritizes speed and margins over model prestige.And this is really another part of the AI race that is built for real use. This is especially important for Google where AI it's not just a standalone product but its AI has to run inside of search ads ...
Google launches Gemini 3 Flash, makes it the default model in the Gemini app
TechCrunch· 2025-12-18 00:00
Google today released its fast and cheap Gemini 3 Flash model, based on the Gemini 3 released last month, looking to steal OpenAI’s thunder. The company is also making this the default model in the Gemini app and AI mode in search.The new Flash model arrives six months after Google announced the Gemini 2.5 Flash model, offering significant improvements. On the benchmark, the Gemini 3 Flash model outperforms its predecessor by a significant margin and matches the performance of other frontier models, like Ge ...
OpenAI神秘生图AI爆出,实测不敌谷歌一根香蕉,网友:就这?
36氪· 2025-12-11 10:50
就在今天,OpenAI两款神秘图像模型,悄然登陆Design Arena和LM Arena平台展开测试。 Chestnut(栗子) Hazelnut(榛子) 这周重头戏即将上演!OpenAI神秘生图模型GPT Image 2内部代码流出,将随GPT-5.2一同上线。另一边,谷歌Flash版「香蕉」和Gemini 3 Flash也准备 好了。 没蹲到GPT-5.2? 别急,OpenAI的「王炸」已经在路上了! 这两款模型,据称是GPT Image 2。有网友猜测,它将随着下一代旗舰模型GPT-5.2于本周亮相。 与谷歌Nano Banana Pro相比,它们具备了类似的世界知识,可生成画质相近的名人自拍,特别擅长在图像中嵌入代码。 下图中,它可以生成手写风格的白板字(但logo好像不是最新的),还能在生图时直出代码。 另有实测过的开发者们,纷纷表示OpenAI生图质量,还是比不上谷歌Nano Banana Pro。 尤其是人物面部生成效果,看起来比较假。 对此,他们认为,GPT Image 2可能仍基于GP-4o完成的训练,不过相较于上一代还是提升了不少。 神秘生图模型上线,打不过Nano Banana? 那么 ...
腾讯研究院AI速递 20251211
腾讯研究院· 2025-12-11 00:01
生成式AI 2. 开发者实测显示OpenAI生图质量尤其在人物面部生成效果上不如谷歌Nano Banana Pro,推测可能仍基于GPT- 4o训练但相较上一代有所提升; 3. 谷歌Nano Banana Flash新模型也将本周登场,同时Gemini 3 Flash即将发布,OpenAI与谷歌正面争锋AI"大 戏"即将上演。 https://mp.weixin.qq.com/s/VsFTC-85bLSvma7icsbb_w 二、Mistral再开源!发布代码模型Devstral 2及自家原生CLI 1. Mistral AI发布下一代代码模型系列Devstral 2(123B)和Devstral Small 2(24B),在SWE-bench Verified 上分别达到72.2%和68.0%,成本效率比Claude Sonnet高出7倍; 2. 同步发布原生CLI工具Mistral Vibe,提供项目感知上下文、智能引用和多文件编排功能,已作为Zed扩展程序提 供; 3. 采用修改版MIT许可证增加收入限制条款,月收入超2000万美元的公司无权免费使用,需联系Mistral AI购买商业 授权。 http ...