Back to News

Google DeepMind releases Nano Banana 2.1 image model based on Gemini 3.6 Flash

#google#image-generation#gemini#nano-banana

Google DeepMind has released Nano Banana 2.1, an efficient image generation and editing model based on Gemini 3.6 Flash, now stable in the Gemini API. The update claims broad improvements in visual design, mask editing, subject consistency, and image naturalness, with pricing about 50% lower than Nano Banana 2. The model supports 1K, 2K, and 4K outputs, handles up to 14 reference images for multi-image fusion, and is distributed across Gemini App, Google AI Studio, Gemini API, Google Search AI Mode, Google Ads, Flow, and Stitch.

Coverage timeline

  1. Techmeme

    Matthias Bastian / The Decoder : Google releases Nano Banana 2.1, based on Gemini 3.6 Flash, saying it improves on previous versions “across the board”; pricing is ~50% lower vs. Nano Banana 2 — Google has released Nano Banana 2.1, a new model for image generation and editing.

  2. 机器之心机器之心

    Nano Banana 2 推出几个月之后,谷歌终于更新了这条产品线。 就在今天凌晨, Google DeepMind 正式发布 Nano Banana 2.1 。 从命名看,这只是一次「0.1」级别的升级,但谷歌给出的评价相当直接:新版在此前 Nano Banana 系列基础上进一步提升,重点变化集中在视觉设计、蒙版编辑、主体一致性和图像自然度。 更值得关注的是,Nano Banana 2.1 并没有改变这条产品线原本的方向。它依然追求 Flash 级别的速度和成本效率,同时继续向专业图像生成和编辑靠近。谷歌将其定义为最新的高效率图像生成与对话式编辑模型。 目前,Nano Banana 2.1 已经成为 Gemini API 中的稳定模型,模型 ID 为 gemini-nano-banana-2.1。按照谷歌公布的信息,它已经分发到 Gemini App、Google AI Studio、Gemini API、Google Search AI Mode、Google Ads、Flow 和 Stitch 等产品与开发平台。 从谷歌公布的模型卡和评测成绩来看,这次提升也并非只体现在几个展示案例上。 模型卡地址:https://storage.googleapis.com/deepmind-media/Model-Cards/Nano-Banana-2-1-Model-Card.pdf 底座换成了 Gemini 3.6 Flash Nano Banana 2.1 属于 Gemini 3 系列,底层基于 Gemini 3.6 Flash。 这意味着,它延续了 Gemini 原生多模态模型的能力,可以理解文本和图像输入,再生成图像与文本。谷歌对它的定位也很明确:面向高效率的图像生成和对话式编辑,在速度和成本上继续保持 Flash 路线,同时提高最终输出质量。 在开发者侧, 新模型支持 1K、2K 和 4K 输出,谷歌也专门优化了 1:4、4:1、1:8、8:1 等宽幅和超宽幅画面的平铺伪影问题 。与此同时,多图融合最高可以接收 14 张参考图,并支持多角色和多物体的一致性处理。 这些更新背后的方向很清楚。谷歌让 Nano Banana 从「生成一张好看的图」,继续走向一套可以反复修改、保持素材一致、处理复杂版式的视觉生产工具。 最明显的提升,首先发生在「设计感」上 生成模型

Related stories

  • MiniMax unveils M3.1 Flash, claiming top agent scores and low cost

    MiniMax released M3.1 Flash, the official version of the previously anonymous Space Bunny model, claiming a generational leap in agentic tasks, coding, computer use, and cybersecurity. The model scores 29.6% on Agents' Last Exam (ALE), the top for Chinese models, with pricing at 1 yuan per million input tokens and 4 yuan per million output tokens. Weights are due October 23, with day-0 support from vLLM on Blackwell, Rubin, and AMD MI350 GPUs.

    4 reports · 4 sources
  • Google releases Nano Banana 2.1 image model with 4K output, mask editing, lower API prices

    Google released Nano Banana 2.1, a new image generation model with upgrades in visual design, mask editing, subject consistency, and realism, supporting 4K direct output and improved Chinese text rendering. According to Arena, it averaged about 61.7 points higher than Nano Banana 2 across text-to-image, multi-image editing, and single-image editing, ranking fourth in text-to-image. Google also cut API prices by half for 1K and 2K image generation and about 25% for 4K.

    1 reports · 1 sources
  • Google launches Gemini 3.8 Live and 3.8 Live Extended Thinking speech-to-speech models

    Google released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, two new speech-to-speech models described as its most advanced live dialogue models yet, aimed at enabling voice agents. The models support real-time voice conversations with interruption capability, and a third-party developer has built a browser-based web UI using the WebSocket endpoint and Web Audio API.

    4 reports · 4 sources486 points · 327 comments
  • Google to restrict free Gemini access to Flash-Lite from Oct 9

    Google will limit free Gemini users to the 3.5 Flash-Lite model starting October 9, removing free access to Flash and Pro models. AI Plus subscribers ($4.99/month) will be restricted to 3.5 Flash-Lite and 3.6 Flash, losing access to Gemini Pro, which will require AI Pro or Ultra tiers along with the Deep Think reasoning option.

    3 reports · 3 sources59 points · 52 comments

News updates every hour. Subscribe in your RSS reader to get new stories automatically.

Subscribe