谷歌再次刷新文生图榜单。 去年8月，谷歌发布了Gemini图像模型Nano Banana，一度全网刷屏，成为现象级产品，同年11月，谷歌又发布了Nano Banana Pro，提供更高级的智能功能和 工作室级别的创意控制。 北京时间2月27日凌晨，谷歌又更新了，这次是Nano Banana 2（Gemini 3.1 Flash Image），兼具了速度和Pro版的性能，同时价格也更便宜了。谷歌表示，这 是团队目前最好的图像生成和编辑模型。 | Current models | All models | All | Open weights | First party foundation models | All | Global Leaderboard | Personal Leader | | --- | --- | --- | --- | --- | --- | --- | --- | | 11 | Creator TJ | | Model 17 | | ELO TT | 95% Cl | Samples TJ | | 1 | G Google | | | Nano Banana 2 (Gemini 3 ...

Core Insights - Google has launched Nano Banana 2, which emphasizes "speedy experience" and "professional image quality," with a significant new feature of "real-time connectivity" that enhances its capabilities beyond mere image generation [1][10]. Group 1: Product Features - Nano Banana 2 integrates with Gemini's search capabilities, allowing the model to understand, retrieve, and generate images that are more aligned with real-world information structures [1]. - The model can generate detailed street scenes and character interactions that are nearly indistinguishable from real photographs, showcasing its advanced rendering capabilities [2][3]. - The "real-time connectivity" feature allows for precise generation of images based on real geographical and meteorological data, enhancing the model's utility in various contexts [5][41]. Group 2: Competitive Landscape - In the latest Artificial Analysis rankings, Nano Banana 2 secured the top position, with its image editing capabilities ranking third, while being priced at half of its closest competitor, OpenAI [8][9]. - The competition in the image generation sector has intensified, with leading models showing minimal score differences, indicating a close race among top players [9]. Group 3: User Experience and Applications - Users have reported that Nano Banana 2's ability to generate high-quality images with accurate text rendering has significant implications for marketing materials and global communication [45]. - The model's enhanced consistency in character design and scene elements allows for seamless storytelling in comics and branding [51]. - The ability to visualize complex concepts and data efficiently positions Nano Banana 2 as a transformative tool in education, research, and data analysis [43][42]. Group 4: Technical Upgrades - The model has improved text rendering and translation capabilities, allowing for natural integration of text within images, which is crucial for marketing and promotional content [45]. - It supports multiple resolutions, including a new 512px option optimized for low-latency scenarios, making it suitable for rapid prototyping and iteration [64]. - The visual quality of generated images has been upgraded, with more natural lighting, richer materials, and sharper details, making it a viable tool for professional use [66].

Model Performance - Imagen 4 Ultra 被认为是目前全球最佳的文本到图像模型 [1] - 该模型正处于大规模生产应用阶段 [1] Availability - Imagen 4 Ultra 现已在 Gemini API 和 AI Studio 中可用 [1]

Product Release - Google is launching Imagen 4 and Imagen 4 Ultra in the Gemini API + Google AI Studio [1] - Imagen 4 is available for free trial in AI Studio and in paid preview in the API [1]

Core Viewpoint - The article discusses the launch of Gemini Robotics On-Device, a new visual-language-action (VLA) model by Google DeepMind, designed for robots to operate efficiently without continuous internet connectivity [1][2]. Group 1: Product Overview - Gemini Robotics On-Device is the first VLA model that can be directly deployed on robots, enhancing their ability to adapt to new tasks and environments [2][4]. - The model is optimized for efficient operation on robotic hardware, showcasing strong general flexibility and task generalization capabilities [4][12]. - It can operate in environments with no data network, making it suitable for latency-sensitive applications [5]. Group 2: Developer Tools - Google will release the Gemini Robotics SDK, allowing developers to evaluate the model's performance in their specific tasks and environments [7]. - Developers can test the model in DeepMind's MuJoCo physics simulator, requiring only 50 to 100 demonstrations to adapt to new tasks [7][21]. Group 3: Performance and Adaptability - Gemini Robotics On-Device has demonstrated strong performance in various dexterous tasks, such as unzipping bags and folding clothes, all executed directly on the robot [12][16]. - The model shows significant advantages over previous local robot models, especially in challenging out-of-distribution tasks and complex multi-step instructions [15][16]. - It can be fine-tuned for improved performance and can adapt to different robotic platforms, including the Franka FR3 and Apollo humanoid robots [25][26]. Group 4: Updates and Changes - Alongside the new model, Google DeepMind has reduced the free usage limits for its Gemini 2.5 Flash and Gemini 2.0 Flash models, which may not be well-received by free users [30][32]. - The company has also announced the launch of new image generation models, Imagen 4 and Imagen 4 Ultra, in its AI Studio and Gemini API [33].