28 November 2025, TechRadar reports that OpenAI and Google have adjusted backend free‑access policies for some of their advanced models. Affected models include Google's Gemini 3 Pro and Nano Banana Pro, as well as OpenAI's Sora. Neither company issued a large public announcement about the changes; instead they are appearing gradually in the user experience: reduced free quotas, certain advanced capabilities moved to paid tiers, and degraded or throttled free requests during peak periods.
The report cites user feedback that models previously accessible for free may now show a "limited free calls remaining" notice after initial access, or begin responding slowly after several complex prompts. Some developers see stricter rate limits and lower concurrency quotas in their consoles; "Switch to Pro" prompts on mobile and web are appearing more frequently than before.
TechRadar notes that this round of adjustments is not a single‑point failure but a strategic tightening by both companies to address "surging demand and sustained high compute costs." Typical changes include noticeably fewer free calls per day or month, moving free support for multimodal inputs (such as images or voice) into paid packages, and restricting extended context lengths or advanced analysis features to paying users.
For regular users, this means workflows that relied on free access for frequent experimentation or daily assistant use may be interrupted. TechRadar advises that if you are suddenly asked to upgrade, first check your account quotas and usage records, as short bursts of high‑concurrency requests can trigger new limits. For creators and content producers who need stable long sessions or larger context windows, paid plans remain the more reliable option.
The impact is greater for developers. The report mentions small teams encountering lowered traffic thresholds during testing, causing CI pipelines, automated tests, or demo systems to be frequently disrupted. TechRadar recommends developers review their API request strategies promptly: reduce unnecessary retries, sample cold‑start requests, run bulk tests during off‑peak hours, or switch to local lightweight models as a backup.
TechRadar's analysis sees the tightening as reflecting an industry shift from "unlimited free trials" to a normalized "free+paid" mix. Companies must balance user growth with operational costs. Short‑term user responses include checking platform announcements and quota pages, considering pay‑as‑you‑go or monthly subscriptions, or looking for open‑source local models (for example lightweight offline inference tools) as alternatives.
In closing, TechRadar advises users to monitor platform policy pages and console notices closely throughout 2025 and adjust usage strategies accordingly. For emerging creators and independent developers, the changes bring cost pressure but also encourage more rational model selection and efficient call design.