# Google DeepMind 发布 Nano Banana 2 Lite 图像生成模型

- 来源：Ars Technica：AI（RSS）
- 作者：Ryan Whitwam
- 发布时间：2026-07-01 02:36
- AIHOT 分数：69
- AIHOT 链接：https://aihot.virxact.com/items/cmr0zn6y800bqsldxkke5cwbw
- 原文链接：https://arstechnica.com/ai/2026/06/googles-new-nano-banana-2-lite-image-model-is-its-fastest-and-cheapest-yet

## AI 摘要

Google DeepMind 发布 Nano Banana 2 Lite（正式名 Gemini 3.1 Flash Lite Image），默认低思考模式下约4秒生成一张图像，标准 Nano Banana 需约20秒。API 价格为每1000张图像 $0.034，输入 token $0.25/1M，输出 token $1.50/1M，是 Nano Banana 2 的一半，输出成本为 Nano Banana Pro 的八分之一。用户 Arena.ai Elo 评分接近非 Lite 版本，但在文本处理、小字和角色一致性上较弱。所有输出图像均带有 SynthID 水印。即日起可通过 Google AI Studio、API 及 Gemini（选择 Flash-Lite 选项）使用。

## 正文

Story text

* Subscribers only
Learn more

There are plenty of AI image-generation models these days, but the ones capable of quality outputs tend to be slow and expensive. Google DeepMind says its new image model, known as Nano Banana 2 Lite, offers the best balance of quality and speed. It’s available today across the Google ecosystem, creating images in a fraction of the time it takes Google’s beefier models.

The new model is part of the Gemini 3.1 family—it’s technically called Gemini 3.1 Flash Lite Image. On one hand, Google says this model is ideal for exploring ideas and “rapid-fire” prototyping, applications in which quality can take a backseat. However, the company has also provided some examples aimed at showing how close Nano Banana 2 Lite can get to the quality of its other image models.

A comparison of Nano Banana 2 Lite with the non-Lite version.

Credit: Google

A comparison of Nano Banana 2 Lite with the non-Lite version. Credit: Google

In addition to the examples, Google also has Elo scores from Arena.ai ready to go, showing that users rate Nano Banana 2 Lite outputs almost as highly as the non-Lite versions. However, vibemarking doesn’t always focus on the details that can make AI images look silly upon closer inspection. Google notes that Nano Banana 2 Lite tends to have more trouble with text, particularly if it’s very small, and infographics are more likely to include incorrect data. Characters and people may also show poor consistency across iterations.

Credit: Google

Credit: Google

But it’s fast. You can go from text to an image in about 4 seconds in the default low-thinking mode. Generating the same images in the standard Nano Banana takes about 20 seconds. The speed and efficiency mean developers accessing Nano Banana 2 Lite via the API will pay a lot less. Google says it averages out to $0.034 per 1K image. The API rates are $0.25 for 1M input tokens and $1.50 for 1M output tokens. That’s half the rate for Nano Banana 2. For Nano Banana Pro, the input tokens are only a little more at $2 for 1M, but the output pricing is $12 (eight times higher).

If you’re just tinkering with an AI model to make a single image, it probably makes sense to use the Flash or Pro versions of Nano Banana. They can handle text better, and anything photorealistic will be more accurate. But for rapid iteration and design inspiration, Nano Banana 2 Lite could save you a lot of time and money.

The rapid-fire nature of Nano Banana 2 Lite also means there may be a lot more AI slop circulating online. Google is not the only company offering cheap image generation, but Nano Banana 2 Lite does appear to be a step up in quality compared to other low-cost options. Google notes that all images produced by Nano Banana 2 Lite have SynthID watermarks, which theoretically labels them as AI even after edits.

You can play with Nano Banana 2 Lite in Google AI Studio, and it’s live in the API. You can also access the new image model in Gemini by selecting the Flash-Lite option and asking for an image. Google says it’s also expanding access to Gemini Omni Flash, which it announced at I/O in May. Omni, which is still limited to video generation, is now available in the Gemini API and Google AI Studio. It previously launched in the Gemini app and Google Flow.

Ryan Whitwam Senior Technology Reporter

Ryan Whitwam is a senior technology reporter at Ars Technica, covering the ways Google, AI, and mobile technology continue to change the world. Over his 20-year career, he's written for Android Police, ExtremeTech, Wirecutter, NY Times, and more. He has reviewed more phones than most people will ever own. You can follow him on Bluesky, where you will see photos of his dozens of mechanical keyboards.

Prev story

1. Sony erases digital content from libraries; we're reminded we don’t own what we buy

2. NASA's X-59 "frankenjet" tests supersonic flight without the sonic boom

3. US renewable boom passes key milestone in April

4. In a bold move, Rocket Lab acquires Iridium Communications

5. US offers $10 million for info on group behind Signal and WhatsApp hacking spree