Google and OpenAI are cutting back free usage for Gemini 3 Pro and Sora 2 — here are the new limits

Bad news for those enjoying consumer AI tools, but not enough to pay for them. In the last few days, both OpenAI and Google have taken steps to reduce free access to their generative AI platforms, perhaps indicating that the ‘free lunch’ period of AI access is beginning to wind down — or at least become far less appealing.

Firstly, 9to5Google spotted a change of wording on Google’s own support page for Gemini 3 Pro. When it arrived last week to great fanfare, free users were promised “up to 5 prompts / day”, matching the limits available on Gemini 2.5 Pro. Free users were also able to use Nano Banana Pro (the model’s image generation element) to create “up to 3 images / day”.

“Thinking with Gemini 3 Pro” no longer has a figure attached. Instead the wording has changed to something far more vague: “Basic access — daily limits may change frequently”. Nano Banana Pro was down to “2 images / day”, though at the time of writing it’s back up to three, with a note saying it’s in high demand and that “limits may change frequently”.

OpenAI, meanwhile, has put similar caps on video generation from its Sora 2 app. Non-paying customers are now limited to six generations a day, down from 30 at launch.

In this instance, we have a clear explanation as to why the change is being made, thanks to a post on X from Bill Peebles, head of Sora at OpenAI: “Our gpus are melting, and we want to let as many people access sora as possible!”.

In other words, Sora 2 is a victim of its own success, with the massive compute requirements needed by AI video generation causing problems for Open AI’s servers.

In both cases, it seems Google and Open AI have decided the time has come to try and push free users into becoming fully paying customers.

Time to pay?

Google is pretty clear in the footnote about usage limits that it will prioritize paid accounts over free ones. “If capacity changes, limits for users without a Pro or Ultra plan may be limited before users with a plan,” it reads, adding that such limits help “maintain a high standard of quality”.

To be clear, there have always been reasons to pay for AI tools if you find them useful. As the chart below shows, it’s not just about getting more prompts. For example, the context window — the amount of data Gemini can digest from an uploaded file — jumps from 32,000 to one million, allowing for far more thorough responses.

Nonetheless, those who have gotten by on free models up until now will not like the direction of travel. Upselling a paid product with a limited free version is a tactic as old as the hills, but reducing the cost-free version’s usefulness is a gamble. Yes, it may drive more people to paid accounts out of frustration, but it’s equally plausible they’ll just find themselves using the product less.

That said, if all free options are becoming less generous — as seems likely given the huge energy and hardware costs for AI businesses looking to keep up — the choice may eventually be a binary one: pay up, or opt out of AI use altogether. For those who are already hooked on the tech’s convenience, the free lunch era may finally be coming to an end.

More from Tom's Guide

• ChatGPT-4o vs. ChatGPT-5.1 — I tested both and the winner surprised me
• I tested Gamma, the AI that builds slide decks in seconds — here’s what impressed me (and what didn’t)
• I tested ChatGPT vs Gemini vs Claude to see which chatbot is the biggest people-pleaser — one went way too far and compared me to Steve Jobs

Follow Tom's Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds.