Skip to content
TopicTracker
来自 simonwillison.net查看原文
译文语言译文语言

Nano Banana 2 Lite

Nano Banana 2 Lite(亦称 Gemini 3.1 Flash Lite Image)是 Google DeepMind 推出的"最快、最便宜的 Gemini 图像模型",专为速度和规模而设计。作者使用 AI Studio 让它生成一张"沃利在哪里"风格但主题为"浣熊拿着业余无线电在哪里"的图片,结果比今年四月尝试的其他 Nano Banana 模型效果更好,不过它将"森林节"拼错了两次。

背景速读

- Simon Willison 是知名技术博主,长期追踪 AI/LLM 领域动态,他的评测文章在开发者社区有较高影响力。 - "Nano Banana" 是 Willison 本人发起的一个非正式测试系列,用"画一张《威利在哪里》风格、但找的是一只拿着业余无线电的浣熊"这个固定提示词,横向比较不同图像生成模型的质量——包括文字渲染、构图能力、逻辑一致性。命名"香蕉"源自硅谷常用的"香蕉标准"(即用香蕉作参照物来直观比较尺寸、性能)。 - Gemini 3.1 Flash Lite Image(文中戏称 Nano Banana 2 Lite)是 Google DeepMind 2026 年 6 月底发布的最新轻量级文生图模型,主打速度快、成本低,适合大规模调用。 - 文中提到的 2026 年 4 月测试是指 OpenAI GPT-Image 2 发布时 Willison 用同一提示词做的对比,结果当时 GPT-Image 2 表现明显优于早期 Gemini 模型。 - 这篇短文的核心观察:Google 的新轻量模型已大幅缩小差距,但文字拼写仍不可靠("Forest Festival" 拼错两次),这是当前所有文生图模型的共性难题。

相关报道

  • Jacob Gold describes using a "Nano Banana Pro" device to view historical satellite imagery and witness changes to Earth's surface over time, offering a unique perspective on environmental and urban development from the past.

  • Google DeepMind released a new lightweight image model called Nano Banana 2 Lite under the Gemini brand. The model is designed for efficient, low-cost image generation tasks, prioritizing speed and reduced computational requirements while maintaining quality.

  • Google has introduced two new AI models: Gemini Omni Flash, a multimodal model designed for real-time, cross-modal reasoning, and Nano Banana 2 Lite, an optimized small model for on-device applications. Both models are now available for developers to start building with, focusing on efficiency and low-latency performance across various tasks.

  • Testing by LTT Labs found that Belkin's USB-C cable does not charge the Nintendo Switch 2 faster than standard cables. Despite marketing claims, real-world charge speeds were comparable to other quality cables, showing no meaningful performance advantage for the Switch 2.