The free tier is being retired

Thirty-five of the 75 products in the second edition of the AI Credit Index have a $0 plan. That number went up this edition, from 26 of 52, mostly because eight of the 24 new products brought free plans with them. So the free tier is not disappearing. What is disappearing is the thing that used to be inside it: a monthly allowance of credits, printed on the pricing page, that refreshes whether or not you ever pay. That shape is being replaced, product by product, with shapes that look free on the card and behave like a trial.

This essay is the record of that replacement across one quarter, and the three shapes it is taking.

What moved#

On the products the Index observed in both June and October 2026, six changed their free tier.

Pika cut its free plan from 80 video credits a month to zero. The card now says “Free 0 credits / month · packs only”, and the FAQ confirms a top-up is available “on any plan, including Free”. The smallest pack is $25. I took that relaunch apart in the Pika essay; for this one the relevant fact is that a free user who used to get 80 credits a month, watermarked and limited to 480p, now gets none, and the first clip costs $25.

Ideogram stopped printing its free allowance. June’s page said 10 credits a week; September’s says the free allowance “may vary”, and the pack sizes were de-published with it. The Index records that as a free tier whose size is no longer public, which for a buyer is the same as no free tier.

v0 converted a monthly $5 of free credit into a one-time $5 for new users. The docs are explicit: “New users who sign up under the one-time Free credit offer receive $5 once per user”, while existing free users keep their monthly $5. The free tier still exists for everyone who already had it, and is a trial for everyone who arrives now.

Canva restated its Free allowance from 200 uses to 20, but the unit changed with it, from any AI tool to Premium-tool uses only, so the Index treats that as a series restart rather than a 90% cut. The honest reading is that the free allowance was redefined in a way that makes the old and new numbers incomparable.

And two went the other way. Atlassian gave Rovo Dev a free tier of 350 credits and moved the Rovo Standard add-on to a free beta. PhantomBuster’s free plan now includes 30 minutes of execution time a month, stated in a vendor blog post “as of July 2026”.

Among the new products, the one clear retirement is Tasklet: its free plan, which carried 300 bonus credits a day on the June 2026 page, was still there on 3 September and gone by 10 October. Chatbase had already halved its free credits from 100 to 50 a month by December 2025, and its free agents are deleted after 14 days of inactivity. Two smaller moves on products observed in both editions: Replit added a “Free Mode” inside its free Starter plan, with daily credits “up to a monthly cap”, and Genspark softened on its paid side, adding a weekly free quota drawn before monthly credits and extending its grandfathered unlimited chat “through December 31, 2026”; its free tier held at 100 credits a day.

Net across the panel: two cuts to zero or to unstated, one monthly-to-one-time conversion, one redefinition, one plan removed, two additions. Direction is mixed; the shape is not.

The three shapes#

Line up the free plans in Edition 2 by how the allowance arrives and the pattern is plain. Of the 32 products that state a free allowance, 19 refresh it monthly. Seven drip it daily. Six grant it once.

The one-time grant. Gamma gives new free users 400 credits at sign-up and caps a free account at 2,000. ClickUp’s free workspace gets 500 AI credits in total, not per month. Fellou’s free plan is “1000 Sparks (New User Bonus)”, roughly four tasks. Descript and Runway have carried one-time grants since Edition 1. v0 joined them this quarter. These are trials with a $0 price tag: the credits run out, the plan does not refill them, and the card can still say Free.

The daily drip. Genspark gives free users 100 credits a day. Recraft and PixVerse give 30. Manus gives 300, Leonardo 150, Suno 50, Udio 10. Figma’s Starter plan allows 150 AI credits a day up to 500 a month. Tasklet’s free plan was 300 a day until it was removed. A daily allowance looks generous summed over a month and is useless for a piece of work that takes more than a day’s worth; it cannot be saved up. It is a throttle on what a free user can accomplish, which is precisely its purpose.

The monthly allowance, shrinking or unstated. This is the shape that is leaving. Nineteen products still state one, and Pika’s is now zero. Chatbase halved its number. Around them, Ideogram stopped stating its allowance at all, and Murf’s free plan lists no recurring minutes, only a one-time trial. HubSpot’s free users “cannot use HubSpot Credits”. The products that added a free monthly allowance this quarter did so for developers (Rovo Dev’s 350 credits) or in a low-cost unit (PhantomBuster’s 30 minutes).

Why now#

On a seat-priced product a free user costs the vendor almost nothing, so free plans were funnels by default. On a credit-metered product every free generation has a cost of goods, and the burn tables that the Index records make that cost legible to the vendor in a way it was not two years ago. On Pika’s June page a 720p clip cost 20 credits; 80 free credits a month was the credit cost of four of them, every month, forever, for anyone who never upgraded. At the volumes a consumer video tool runs, that is a real line in the P&L.

So the question each vendor is answering is not “should we have a free plan” but “what is the free plan for”. If it is for acquisition, a one-time grant does the job and then stops costing money. If it is for retention of light users who might upgrade later, a daily drip keeps them in the product without letting them do real work on it. If the product’s economics cannot support either, the number is set to zero and the free plan becomes a storefront for packs, which is where Pika landed. The only reason left to keep a real monthly allowance is competition for a user base you have not won yet, which is why the additions this quarter are in developer tools and automation, where the fight for adoption is still on.

Should you meter at all? argued that a meter changes what every tier is for. The free tier is the first place that shows up, because it is the tier where the vendor’s cost is not offset by any revenue at all.

What to do with this#

If you rely on a free AI plan, read the cadence, not the headline. A daily drip is a cap on a day’s work. A one-time grant is a trial. A monthly allowance can be set to zero between two captures, as Pika’s was, and “may vary” means the vendor has stopped committing to a number.

If you are deciding whether to pay, price the free plan at what it would cost you on the first paid tier. Pika’s old free allowance was worth about 90 cents a month at Starter rates; v0’s one-time $5 is $5. The vendors have done this arithmetic; most of the free plans that survived it are the ones worth least.

If you run a credit-metered product, the panel suggests the order of operations: convert the monthly allowance to a one-time grant before you cut it, drip daily if you need free users in the product, and if you go to zero, publish the pack price next to it so the free plan is honestly a storefront. The one thing not to do is what Ideogram did, which is to keep the free plan and stop saying what it contains.

Every free-tier figure here is on the product’s page in the Index with the archived capture it came from, and each page states the cadence; the methodology explains how a free allowance is recorded.