The Gemini 3 Pro family
Gemini 3.1 Pro does not have a price, it has eight: four service tiers, each splitting at the 200k token line. The cheapest case is $1 in and $6 out and the dearest is $7.20 and $32.40, for the same tokens.
- four service tiers
- the 200k token step
- context window unpublished
Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. What changed
Eight numbers that all belong to one model
The Google pricing page carries four separate tiers for this model. Standard is $2 in and $12 out below 200k tokens, and $4 and $18 above it. Batch is exactly half of standard: $1 and $6, moving to $2 and $9 past the line. The Flex tier carries the same numbers as Batch. Priority is $3.60 and $21.60, going to $7.20 and $32.40 past the line.
Cheapest to dearest, for a long prompt, is a 3.6 times spread. None of those eight numbers is wrong, and none of them alone is the price of Gemini 3.1 Pro. If you have seen a single figure quoted for this model, that source did not tell you which tier and which side of the line it read.
The 200k line is your most expensive architecture decision
Crossing 200k tokens doubles the input price and multiplies output by one and a half. So if you are building something that resends an entire conversation history or a whole repository on every call, the day your input crosses that line the invoice steps up while nothing in your code has changed.
The practical move is to count the input length before sending and keep it under the line where you can. This is the kind of optimisation that is invisible in the unit price and very visible in next month invoice.
The number Google does not give
The Google models page prints no context window for gemini-3.1-pro-preview. All four Chinese vendors in this reference publish that figure, Alibaba goes further and breaks out maximum input and a chain-of-thought ceiling, and Google withholds the simplest one.
The consequence for our table is direct: the context column for this candidate stays empty and scores nothing. We do not fill that cell with a plausible number, because an unsourced fact in a comparison table is worse than an empty cell. The odd part is that the pricing announces a step at 200k tokens while never stating the overall ceiling.
The family name and the model that actually exists
This family is named 3 Pro and its only live text model is 3.1 Pro, still in preview. No model called gemini-3-pro-preview appears on the Google pricing page. The only thing still carrying the 3 Pro name is an image model, gemini-3-pro-image, whose status is stable and whose public name is Nano Banana Pro.
So if you arrived looking for Gemini 3 Pro, what you can buy is either 3.1 Pro in preview or an image generation model. That image model takes its text pricing from 3.1 Pro, and its image output runs $120 per million tokens, which works out to roughly $0.134 for a 1K or 2K image and $0.24 for a 4K one.
Google Search inside the model, and a quota that is shared
There is one pricing row only Google has: grounding with Google Search. The first 5,000 search requests a month are free, then it is $14 per 1,000 requests. The detail that slips past people is that those 5,000 are shared across all Gemini 3.x models rather than being this model own allowance. If you run several services on several Gemini models, they all draw from one pot.
Release timeline
- Gemini 3.1 Pro preview Tiered pricing and a preview ID; Google publishes no context window for it.
Using it from Iran
This column carries a date and says how it was checked, because most listicles guess it. Where we have only read a vendor policy page, the note below says exactly that.
- Reachable
- blocked
- Payment
- no working route
- Free tier
- no
- Measured on
How we checked: The Gemini API available-regions page lists the countries where the API works, and Iran is not on that list. Read there rather than measured on the network.
What it is good at
- The Batch tier is exactly half of Standard, so anything without urgency costs half
- Native grounding with Google Search, with 5,000 free search requests a month
- Four service tiers mean cost and priority can be tuned independently
- Context caching is priced separately at $0.20 below the 200k line
Where it falls short
- Google publishes no context window for this model, so the context column in our table stays empty and scores nothing.
- The only text model in this family is still in preview, so neither its behaviour nor its price carries a stability guarantee.
- The price steps at 200k tokens and input doubles, so the cost of a long prompt is not linear.
- There is no free tier; the pricing page lists none for this model.
- The 5,000 free search quota is shared across all Gemini 3.x models rather than belonging to this one.
- The family is named 3 Pro while the available text model is 3.1 Pro, and the only thing still carrying the old name is an image model.
- Iran is not on the Gemini API available-regions list, so there is no official route.
Our take
Before comparing this model against any other, first decide which service tier you are buying, because cheapest to dearest is a 3.6 times spread and most comparisons ignore it. And if the context window is your deciding criterion, this model has no published answer for you today.
Questions people actually ask
What does Gemini 3.1 Pro cost
It depends on the service tier and the prompt length. Standard is $2 and $12 below 200k tokens and $4 and $18 above. Batch is half of Standard and Priority is 1.8 times it. There is no single figure for this model.
What is the Gemini 3.1 Pro context window
Google prints no figure for it on the Gemini API models page. We do not guess, so that cell is empty in our table. The same pricing announces a step at 200k tokens, but that is a price boundary and not the window ceiling.
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