Google Didn't End Free Gemini. It Moved the Free Line Down One Model.

The headline version of this story is that Google pulls free Gemini access. I read Google’s own pages on 14 October to see how much of that holds. Some of it does, with dates. The larger claim, that free Flash and Pro access are gone, is not what the pages say today. The accurate version is a line moving down the model ladder, in two places that follow different rules: the developer API and the consumer Gemini app.

I am anchoring on Google’s documentation, release notes and help pages. Where a news outlet described a page that has since read differently, I say so.

The API: what the pricing page lists

The Gemini API pricing page (last updated 14 October UTC when I loaded it) has a “Free Tier” column per model. Among the eight text models I picked, as I read it in a browser, this is what the column says:

The free column also carries a price that doesn’t show as a number: the same page marks “Used to improve our products” as Yes on the free tier and No on the paid one. Free here means free of charge, with your prompts available for product improvement.

Free-tier listing for eight Gemini API models, 14 October 2026 Eight models in rows. Filled circles mean the pricing page lists the free tier as free of charge: Gemini 3.8 Flash, 3.6 Flash, 3.5 Flash-Lite, 2.5 Pro, 2.5 Flash and 2.5 Flash-Lite. Open rings mean the free tier is not available: Gemini 4 Argon and Gemini 3.1 Pro Preview. FREE TIER ON THE GEMINI API PRICING PAGE · READ 14 OCTOBER 2026 Gemini 4 Argon Gemini 3.1 Pro Preview Gemini 3.8 Flash Gemini 3.6 Flash Gemini 3.5 Flash-Lite Gemini 2.5 Pro Gemini 2.5 Flash Gemini 2.5 Flash-Lite Not available on the free tier — Gemini API pricing page, 14 October 2026 Not available on the free tier — Gemini API pricing page, 14 October 2026 Free of charge — Gemini API pricing page, 14 October 2026 Free of charge — Gemini API pricing page, 14 October 2026 Free of charge — Gemini API pricing page, 14 October 2026 Free of charge — Gemini API pricing page, 14 October 2026 Free of charge — Gemini API pricing page, 14 October 2026 Free of charge — Gemini API pricing page, 14 October 2026 not available not available free of charge free of charge free of charge free of charge, but see the 18 September access note free of charge, but see the 18 September access note free of charge, but see the 18 September access note Filled = listed free of charge. Ring = not available. A listing is not a quota guarantee; see the rate-limits discussion.
Of these eight text models, the free column skips the flagship and the Pro preview. The other six, including 2.5 Pro, are listed as free of charge.

18 September: access to the 2.5 models

The release notes have the one entry that reads as a restriction on existing models. Dated 18 September, it says Google is “limiting access to the 2.5 models to users who have actively used them in the past.” The models are not deprecated, and will be served “until further notice.” New projects are pointed to 3.5 Flash-Lite or 3.8 Flash, with the stated reason that this “helps us maintain sufficient capacity.”

That is a gate on who can start using them, and it sits awkwardly beside a pricing page that still prints “Free of charge” in the 2.5 rows. I can’t tell from the two pages whether a new project’s free key reaches 2.5 Pro. What I can say is that the listing and the access note both exist on Google’s pages as of 14 October, and they describe different things: a price and a permission.

30 September to 13 October: the flagship arrives behind a paywall

Google’s Gemini 4 Argon post carries a 30 September publication date in its page metadata. It says Argon was first going to trusted cyber defenders and that wider release would start “with paid API customers and Google AI Ultra subscribers.” On 13 October the API release notes list Argon as generally available, and the Gemini app release notes say it began rolling out the same day to Google AI Pro and Ultra subscribers. So the app audience widened from the blog’s wording (Ultra) to Pro and Ultra within two weeks, and free users were in neither as of 13 October.

The introductory API price is $2.00 per million input tokens and $10.00 per million output tokens through 30 November, then $4.00 and $20.00 from 1 December, per the pricing page and the latest-model guide. At least the page states the date.

The same pattern shows up one rung down, on the paid side. Gemini 3.8 Flash lists $0.75 input and $3.75 output through 31 December, rising to $1.50 and $7.50 from 1 January 2027; 3.6 Flash lists the same figures. The 13 August release note for Gemini 3.7 Flash also says it was “available at an introductory price through December 31, 2026.” On 8 October, 3.7 Flash and 3.5 Flash were deprecated and their requests automatically routed to 3.8 Flash and 3.6 Flash. All of that is paid-tier pricing. It matters here because the free column sits in the same table. I have no earlier snapshot of that column, so I can’t say whether it changed.

Scheduled doubling of introductory paid API prices, dollars per million tokens Four rows on a linear axis from 0 to 20 dollars. Gemini 4 Argon input goes from 2 to 4 dollars on 1 December 2026. Argon output goes from 10 to 20 dollars on 1 December 2026. Gemini 3.8 Flash input goes from 0.75 to 1.50 dollars on 1 January 2027. Gemini 3.8 Flash output goes from 3.75 to 7.50 dollars on 1 January 2027. Rings are the introductory price and filled circles the later price. PAID API PRICE PER MILLION TOKENS · INTRODUCTORY (RING) TO SCHEDULED (DOT) $0 $5 $10 $15 $20 Argon input, from 1 Dec Argon output, from 1 Dec 3.8 Flash input, from 1 Jan 3.8 Flash output, from 1 Jan $2.00 — Argon input, introductory, through 30 November 2026 $10.00 — Argon output, introductory, through 30 November 2026 $0.75 — 3.8 Flash input, introductory, through 31 December 2026 $3.75 — 3.8 Flash output, introductory, through 31 December 2026 $4.00 — Argon input from 1 December 2026 $20.00 — Argon output from 1 December 2026 $1.50 — 3.8 Flash input from 1 January 2027 $7.50 — 3.8 Flash output from 1 January 2027 $2 to $4 $10 to $20 $0.75 to $1.50 $3.75 to $7.50 Every row exactly doubles. Source: Google's Gemini API pricing page, read 14 October 2026.
Paid prices step up on stated dates. None of this touches the free column, where the flagship is not listed as of 14 October.

The app: two tables for one change

The consumer side has a separate set of rules. Google’s help page on model access says changes to Gemini Apps availability begin on 9 October for users without a Google AI subscription; AI Plus subscribers are to get an email saying when theirs take effect. As I read it on 14 October, the page’s table gives AI Plus Flash-Lite and Flash but not Pro, and AI Pro and AI Ultra all three. For users without a plan the row reads “Auto.” Google’s text explains: most prompts go to Flash-Lite, and “[p]rompts that need deeper reasoning may go to Flash or Pro.” With smart selection off, everything goes to Flash-Lite.

9to5Google’s 3 October report reproduced a table that differs from the one I read on 14 October: no-plan users at Flash-Lite only, Flash and Pro marked unavailable. That reading is the one that spread as free users losing Flash and Pro. I can’t tell from outside whether Google edited the page between 3 and 14 October, so I only state the two readings. If the current page is right, the fair summary is that free users keep a path to the bigger models through auto-routing, with the default weighted toward the smallest. That is a smaller restriction than the headline, and it is still a restriction.

The compute-based usage limits are older. The same page dates them to 17 May for users over 18 and 24 July for under-18 accounts. Its sibling limits page puts the free app at “standard” limits, a 32,000-token context window, against 128,000 for AI Plus and one million for AI Pro and Ultra. As of 14 October it publishes no prompt count.

What the free API tier means in practice

The rate-limits page does something that makes any “free got worse” claim hard to measure: it doesn’t print the per-model free numbers. It says to view active limits in AI Studio, and adds that “specified rate limits are not guaranteed and actual capacity may vary.” What it does print is the tier ladder. The free tier needs an “active project or free trial”; Tier 1 needs a linked billing account; spend-based limits ($10, $50 and $200 per rolling ten minutes for Tiers 1 to 3) show “N/A” for the free tier.

One detail there is easy to miss. Limits are applied “per project, not per API key,” and daily request quotas reset at midnight Pacific time. Minting a second key doesn’t add quota.

Put together, the stated position is this. The free tier is a capacity-managed offering, with no published quota to hold Google to, subject to the 18 September access gate on the older models and with the flagship excluded. “Free” on the pricing page is accurate. The quota that would make it useful sits in a dashboard I can’t read from outside.

Dated Google changes to Gemini access, 16 July to 14 October 2026 A timeline from 16 July to 14 October 2026. 21 July: Gemini 3.6 Flash generally available. 13 August: Gemini 3.7 Flash generally available at an introductory price. 2 September: Gemini 3.8 Flash generally available. 18 September: access to the 2.5 models limited to prior users. 30 September: Gemini 4 Argon announced, first to trusted cyber defenders. 9 October: Gemini app model changes begin for users without a plan. 13 October: Gemini 4 Argon generally available on the API and rolling out to AI Pro and Ultra in the app. GOOGLE'S OWN DATES · 16 JULY TO 14 OCTOBER 2026 21 July — Gemini 3.6 Flash GA, API release notes 13 August — Gemini 3.7 Flash GA at introductory price, API release notes 2 September — Gemini 3.8 Flash GA, API release notes 18 September — 2.5 models limited to prior users, API release notes 30 September — Gemini 4 Argon announced, Google blog 9 October — Gemini app model changes begin for users without a plan, Gemini Apps Help 13 October — Argon GA on the API and rolling out to AI Pro and Ultra, release notes 21 Jul · 3.6 Flash GA 2 Sep · 3.8 Flash GA 30 Sep · Argon announced, cyber defenders first 13 Oct · Argon on paid API, app Pro and Ultra 13 Aug · 3.7 Flash GA, introductory price 18 Sep · 2.5 models: prior users only 9 Oct · app plan changes begin Dates are those on Google's release notes, blog metadata and help page. The 9 October app change reaches AI Plus subscribers on a date set by email.
Three of the seven entries narrow access: the 18 September gate, the Argon launch order, and the 9 October app change. The rest are model releases.

Who it hits, and who it doesn’t

By Google’s published text, the groups affected are:

  1. Free API developers who wanted the flagship. Argon is paid-only, and so is the 3.1 Pro Preview row.
  2. New projects that wanted the 2.5 models. They are limited to prior users, per the 18 September note.
  3. Consumer app users without a plan. Their default is Flash-Lite, with auto-routing to larger models as the pages describe it today.
  4. AI Plus subscribers, who per the help page lose Pro on a date they are told by email.
  5. Paid developers on the introductory prices, who face scheduled doubling. That is a price change rather than a free-tier change, and I list it so the totals are honest.

The group I found no change for, as of 14 October 2026, is free API developers using 3.8 Flash, 3.6 Flash or 3.5 Flash-Lite. Google’s pricing page lists them as free of charge, and it publishes nothing that would let me test how generous that is. If someone has seen those quotas move, the AI Studio dashboard is the primary source, and I would take a dated screenshot of it over any article.

Five days earlier I wrote about a model whose output is free by design and whose usage numbers stood in for revenue. The Google case is the mirror image: the price is on the page, and the quota behind “free” is not.

References

  1. Google (2026). Gemini Developer API pricing. Google AI for Developers; page last updated 14 October 2026 UTC.
  2. Google (2026). Release notes. Gemini API; entries of 21 July, 13 August, 2 September, 18 September, 8 October and 13 October 2026.
  3. Google (2026). Rate limits. Gemini API; page last updated 13 October 2026 UTC.
  4. Google (2026). Latest Gemini models. Gemini API guide; introductory Argon pricing.
  5. Kavukcuoglu, K. (2026). Gemini 4 Argon: our next era of frontier intelligence. Google blog; publication date 30 September 2026 in page metadata.
  6. Google (2026). Gemini Apps’ release updates and improvements. Entry of 13 October 2026.
  7. Google (2026). Changes to Gemini model access and limits. Gemini Apps Help; read 14 October 2026.
  8. Google (2026). Gemini Apps limits and feature availability. Gemini Apps Help; read 14 October 2026.
  9. Li, A. (2026). Gemini app limiting what models free and AI Plus users can access, AI Pro adding Deep Think. 9to5Google, 3 October 2026.