The technical map of Google AI infrastructure
Google runs the widest infrastructure on the market and is the only place where text, image, video, speech, music and a study tool — all six — are unified under one account. The cost of that breadth surfaces at the naming layer: four image units all carry the brand name "Nano Banana" while each carries a distinct technical identifier.
- all six categories covered
- NotebookLM
- naming complexity
Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. Change log
The state of this maker's versions
- 10
- 4
- current 10
- preview 4
These are the versions recorded in this reference, not the maker's whole catalogue.
Products
From Gemini to NotebookLM. Each unit runs its own independent function.
Model families
The model lines, from Gemini Pro through Nano Banana, Veo and Lyria.
Versions
| Versions | Deployed | Status | Model families |
|---|---|---|---|
| Gemini 2.5 Flash | current | Gemini 2.5 | |
| Gemini 2.5 Pro | current | Gemini 2.5 | |
| Gemini 3.1 Flash-Lite | current | Gemini 3 Flash | |
| Gemini 3.1 Flash TTS | preview | ||
| Gemini 3.1 Pro | preview | Gemini 3 Pro | |
| Gemini 3.5 Flash-Lite | current | Gemini 3 Flash | |
| Gemini 3.5 Flash | current | Gemini 3 Flash | |
| Gemini 3.6 Flash | current | Gemini 3 Flash | |
| Gemini 3.7 Flash | current | Gemini 3 Flash | |
| Lyria 3 Pro | preview | Lyria | |
| Nano Banana 2 Lite | current | Nano Banana | |
| Nano Banana 2 | current | Nano Banana | |
| Nano Banana Pro | current | Nano Banana | |
| Veo 3.1 | preview | Veo |
Infrastructure breadth, both an advantage and a source of complexity
In the current Gemini API documentation, text units span Gemini 2.5 through 3.6 Flash, image carries four units, video three, audio six and music three. No other infrastructure covers this range. For multi-modal workloads, one Google account is sufficient, which in practice reduces integration time.
The cost of that coverage surfaces at the naming layer. "Nano Banana Pro" carries the technical identifier gemini-3-pro-image, and "Nano Banana 2" corresponds to gemini-3.1-flash-image. The brand name and the technical identifier do not align, and selecting by brand name produces a configuration error. Reports in this section always log both identifiers together.
Where this infrastructure has no direct peer
NotebookLM has no equivalent-tier competitor. It is a study tool, not a chatbot, and it runs a function no other infrastructure ships as a product: it converts a user own sources into several output formats to simplify information absorption. Search volume confirms that position, running ahead of many more widely known names.
A pricing parameter that gets lost in the rate table
Gemini 3.1 Pro pricing is tiered: up to a 200K-token prompt ceiling, the input rate is $2 and output $12; above that ceiling, the rate moves to $4 and $18. For a service sending large prompts, the real operating rate is the second tier, not the figure shown in the headline. Our data file logs only the first-tier figure, which is exactly why this clarification is logged in this report.
Documented strengths
- The only maker shipping operational units in all six categories: text, image, video, speech, music and study
- NotebookLM has no equivalent-tier competitor and is engineered for processing learning material from a user own sources
- A genuine free tier on AI Studio, sufficient for initial evaluation
- A speech-to-speech translation unit covering more than seventy languages
Known weaknesses and limits
- Brand names and technical identifiers do not align, which drives unit misselection.
- Pro unit pricing is tiered and doubles above a 200K-token prompt ceiling.
- Model documentation does not log context window figures, so that parameter is blank in our table for every Google unit.
- Iran is not logged on the available-regions list. Google routes readers outside that list to Google Cloud, which is not a simple route.
Technical verdict
For multi-modal workloads, Google is the least friction-prone infrastructure because every unit is unified under one account. For workloads limited to coding, that breadth buys nothing and Anthropic remains ahead.
Frequently raised questions, with documented answers
What is the difference between Nano Banana Pro and Nano Banana 2
Nano Banana Pro carries the identifier gemini-3-pro-image, engineered for design work and 4K output, at a $2 input rate. Nano Banana 2 is gemini-3.1-flash-image, built for low-cost production volume, at a $0.50 input rate.
Is Gemini reachable from Iran
Iran is not logged on the available-regions list for the Gemini API or AI Studio. This parameter is extracted from Google own document. No independent network test has been run, and when one is, the date will be logged here.