Google's Gemini 3 Pro, Nano Banana Pro, and Antigravity

Google casually dropped three products — upgraded versions of two of their most popular products, Gemini and Nano Banana, and Antigravity — an IDE with agentic capabilities (a VSCode fork using Windsurf tech since Windsurf founders signed a $2.4 billion deal with Google and joined Google Deepmind along with some other Windsurf employees).

Gemini 3 Pro

According to benchmarks and Google, Gemini 3 Pro is the new state of the art model outperforming other AI models. This brings us back to the Gemini era in the cycle of SOTA (State-Of-The-Art) models. Not sure which model will outperform Gemini 3 Pro in the coming months (or maybe weeks), but for now Gemini 3 Pro seems to be the king.

From what I saw and experienced, Gemini 3 Pro is really good at one-shot prompts and does a great job even if you don't provide it a lot of context. At the very least, it makes things functional and usable.

People are sharing what they built using Gemini 3 Pro on X (Twitter). Here's what they are building:

Nano Banana Pro

"Nano Banana Pro" a.k.a "Gemini 3 Pro Image" is an AI model built on Gemini 3 Pro, and that allows Nano Banana Pro to utilize Gemini 3 Pro's reasoning and thinking capabilities.

To use Nano Banana Pro, go to gemini.google.com (or the Gemini App on your smartphone), click on the tools icon besides the plus(+) icon, and select "🍌 Create Images" option. Also, make sure you are on the "Thinking with 3 Pro" option in the model selection dropdown.

To me, the most amazing thing about Nano Banana Pro is its ability to generate accurate text. And the reason it is able to do so is Gemini 3 Pro, the AI model its built up on. There are countless examples on the internet depicting the amazing ability of Nano Banana Pro to generate infographics, diagrams, banners, menus, etc. all with nearly accurate text.

Google’s Nano Banana Pro is by far the best image generation AI out there.

I gave it a picture of a question and it solved it correctly in my actual handwriting.

Students are going to love this.

— sid on X

In fact, the first image you saw in this post is generated by Nano Banana Pro.

I did notice one thing though, it doesn't generate accurate readable text if the text is in the backdrop/background. I was iterating upon the first image you saw in this post and prompted Nano Banana to generate a slightly modified image with this prompt: "add some texts reflecting on the fragility and hype of these SOTA models and companies, don't overdo it, search the web for grounding the punch lines", and it generated some gibberish text in newspapers in the background.

Also, you can see repetitive text in the heading.

Another thing which I noticed is when you prompt Nano Banana Pro to iterate upon its own generated image, it actually degrades the quality of the image and you can clearly see that the image has been iterated upon. See the two images below.

The image on the left is a result of multiple iterations on the same image internally, while the image on the right is a new image generated from the same image (downloaded and then attached to the chat) with a prompt shown below.

take this image, and generate a new image out of it, don't iterate upon the same image, generate a new 4K image out of it

Antigravity

Antigravity is a VSCode fork using Windsurf technology developed by Google. It's an IDE with agentic capabilities, and browser access and preview support.

You can use Gemini 3 Pro for free for now in Antigravity, but it's highly rate-limited. For me, the usage quota (for Gemini 3 Pro High) gets exhausted pretty quickly within a single agentic session. It's also quite buggy at this point, but we can ignore that as it will definitely improve in the near future.

Antigravity is a direct competitor to Cursor, and it will be interesting to see which of the two gets the edge over the other.

Exciting times.