What Google has in AI
Google has the widest catalogue on the market and is the only place where text, image, video, speech, music and a study tool all sit behind one account. The price of that breadth is naming chaos: four image models all answer to "Nano Banana" while their technical IDs say something else entirely.
- all six categories
- NotebookLM
- naming chaos
Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. What changed
Products
From Gemini to NotebookLM. Each does its own job.
Model families
The model lines, from Gemini Pro through Nano Banana, Veo and Lyria.
Versions
| Versions | Released | Status | Model families |
|---|---|---|---|
| Gemini 3.1 Pro | preview | Gemini 3 Pro | |
| Veo 3.1 | preview | Veo |
Breadth, which is both the advantage and the headache
In the current Gemini API docs the text models run from Gemini 2.5 up to 3.6 Flash, image has four models, video three, audio six and music three. No other lab has that spread. If your work needs several media together, one Google account covers it, and in practice that saves real time.
The cost is the naming. "Nano Banana Pro" is actually gemini-3-pro-image, and "Nano Banana 2" means gemini-3.1-flash-image. The marketing name and the technical ID do not line up, and anyone choosing by name picks wrong. On these pages we always print both.
Where Google stands alone
NotebookLM has no direct competitor. It is a study tool rather than a chatbot, and it does something no other model offers as a product: it turns your own sources into several output formats so the material lands more easily. The search volume backs that up, and it runs ahead of many far more famous names.
A pricing detail that gets lost in the price table
Gemini 3.1 Pro is tiered: up to a 200k token prompt it is $2 in and $12 out; above that it is $4 and $18. If you run a service that sends large prompts, your real price is the second tier, not the number in the headline. Our data file records the first tier, which is exactly why it is written out here.
What it is good at
- The only maker shipping models in all six categories: text, image, video, speech, music and study
- NotebookLM has no peer, and it is built for learning from the user own sources
- A genuine free tier on AI Studio, which is enough for evaluation
- A speech to speech translation model covering more than seventy languages
Where it falls short
- The marketing names and the technical IDs do not match, which makes people pick the wrong model.
- Pro pricing is tiered and doubles above a 200k token prompt.
- The model docs do not print context windows, so that figure is empty in our table for every Google model.
- Iran is not on the available-regions list. Google itself points readers outside the list at Google Cloud, which is not a simple route.
Our take
If your work spans several media, Google is the least painful choice because it all sits behind one account. If you only write code, the breadth buys you nothing and Anthropic is ahead.
Questions people actually ask
What is the difference between Nano Banana Pro and Nano Banana 2
Nano Banana Pro is gemini-3-pro-image, presented for design work and 4K output, at $2 input. Nano Banana 2 is gemini-3.1-flash-image, built for cheap production volume, at $0.50 input.
Does Gemini work from Iran
Iran is not on the available-regions list for the Gemini API or AI Studio. We read that on Google own page. We have not run an independent network test, and when we do the date will appear here.
Sources
- Gemini API models vendor source 12 August 2026
- Gemini API pricing vendor source 12 August 2026
- Gemini API available regions vendor source 12 August 2026