AI Design Tools Pricing: Figma, Adobe and Canva Meters, Priced Per Task

Figma, Adobe and Canva all meter generative AI now, and no two of them count the same thing. A Figma Professional seat gets 3,000 AI credits a month, Adobe Firefly Standard sells 2,000 credits for $9.99, and Canva counts "uses" across three quality tiers. AI design tools pricing only becomes comparable when you price the task, not the credit, so that is what this post does.

The short answer. Price the job, not the credit. For still images, Adobe's Firefly model is unlimited on every paid Firefly plan. Figma charges 2 to 16 credits per image, and Canva draws one Premium or Ultra use. For video and app generation, the three tools diverge so far that only a per-task table gives you a comparable number.

Three AI design tools, three different units of account

The word "credit" suggests a currency you could convert. It is not one. Each vendor invented its own unit, attached it to its own plan structure, and chose which actions sit on the meter and which stay free.

The definitions below are quoted from the vendors' own pages.

Figma and Adobe wording is quoted from pages opened on 10 October 2026. Canva's tier wording comes from its help centre as indexed by search, one step removed from the primary source.

The structural difference is who owns the allowance. Figma gives it to the person: a Full seat gets 3,000 credits on Professional, 3,500 on Organization and 4,250 on Enterprise, per Figma's guide to managing AI credits. Dev, Collab and View seats get 500 each. Adobe sells credits as a product in their own right, with no design software required. Canva never says "credit" at all. The general shape of this problem is covered in the breakdown of why AI billing in credits and tokens confuses buyers.

Figma AI credits: 3,000 per Professional seat, and no public overage price

Figma's rate table is the most granular of the three. It prices each AI action separately and states the rates as of 6 October 2026. Some actions are free: AI search and renaming layers cost nothing.

Image work is cheap. Removing a background costs 1 to 5 credits, vectorising an image 2 to 5, and erasing, isolating, expanding or upscaling 5 to 10. Generating or editing an image costs 2, 6, 8 or 16 credits depending on whether you pick ChatGPT Images 2.0, Nano Banana 2 Lite, Nano Banana or Nano Banana 2.

Agentic work is where the meter runs. Figma's own examples put generating an app from scratch in Figma Make at approximately 100+ credits, and making an attached design interactive at approximately 75+. Asking the Figma agent to generate a 5-screen mobile checkout flow costs 300+ credits. The plus sign matters, because these are floors, not prices.

The seat cap binds before the team pool

Credits are consumed in a fixed order. Seat credits go first, then any shared subscription credits an admin has bought, then pay-as-you-go if it is switched on. A designer who burns through 3,000 credits in week two is blocked unless someone upstream has already paid for more.

Figma began enforcing the per-seat limits on Full seats on 18 March 2026, according to its help centre and support staff on its community forum. Usage earlier in the billing cycle counted against the limit, so some seats hit the wall on day one of enforcement.

Figma does not publish a dollar price for extra credits. The pay-as-you-go rate is visible only inside the admin dashboard's "Add credits" flow. The help centre says only that "the price per credit decreases as more credits are purchased". We think that is the single biggest weakness in Figma's pricing, because a buyer cannot model overage before signing.

What changed on 6 October 2026

Four days before this post, Figma's agent in Design and its Weave tools moved from open beta to general availability. Figma's AI credit updates FAQ states plainly that these features "only started consuming AI credits once they became generally available on October 6, 2026". Weave tools cost from 13 credits for image-to-vector up to 110 for generating product variations.

Figma also softened the blow earlier. From 25 August 2026, add-on purchases gave 2x more credits on Professional and 1.6x more on Organization and Enterprise at the same cost. That is a real price cut on the add-on, and it arrived six weeks before the most credit-hungry features went onto the meter.

Adobe Firefly credits: an unlimited house model, metered partners

Adobe made a structural choice the other two did not. Every paid Firefly plan includes "unlimited standard image and vector generations". The plans page lists Generative Fill, Generative Expand and Text to Vector Graphics among the unlimited standard features. Credits are spent on premium work: video, audio and outputs from partner models.

So the cheapest image on any of the three platforms is a Firefly-model image on a paid Adobe plan, at zero credits. The moment you switch to a partner model, the meter starts.

Where Firefly credits actually go

Adobe's partner-model page prices Gemini 2.5 Flash Image (Nano Banana) at 10 credits per generation, Gemini 3 with Nano Banana Pro at 40, and GPT Image at 60. Generative Fill in Photoshop using partner models costs 10 to 40 credits depending on output size.

Video is the heavy user. The plans page says Firefly Standard's 2,000 credits generate "up to 20" five-second videos, Pro's 4,000 up to 40, and Pro Plus's 10,000 up to 100. That works out to roughly 100 credits per five-second clip at the entry quality setting. Higher resolutions cost more per second.

The add-on price is the only public per-credit rate

Adobe's generative credits page lists four monthly add-on plans: 2,000 credits for $9.99, 7,000 for $29.99, 10,000 for $49.99 and 50,000 for $199.99. Divide through and a credit costs about $0.0050 at the smallest tier and $0.0040 at the largest. No other vendor in this comparison gives you a number you can put in a spreadsheet before you buy.

When credits run out, Adobe says "you still have unlimited access to standard generative features," and it "will not charge you for additional credits unless you choose to purchase them". That is a soft wall rather than a hard one. We think it is the most buyer-friendly design of the three.

Canva pricing in 2026: a meter with no credit and a moving price

Canva is the hardest of the three to price, for two separate reasons. Its unit is a tiered "use" rather than a credit, and its list price has moved repeatedly this year.

Standard, Premium and Ultra tiers

Canva's help centre article on AI usage sorts tools by tier. Premium AI covers Magic Layers, design and image generation with Canva AI, Canva Code interactives, element generation and the music and sound effect generators. Ultra AI covers conversational design with Canva AI 2.0, video clip creation and top-quality image creation. Ultra is available on paid plans only.

On Pro, third-party trackers consistently report an allowance of up to 200 Premium uses or 20 Ultra uses a month from one shared pool. That implies one Ultra use weighs about as much as 10 Premium uses. Canva describes Pro as getting 10x the Free plan's AI and Business 20x, according to PricingSaaS snapshots of Canva's pricing page.

The price moved twice in six weeks

PricingSaaS, which snapshots pricing pages weekly, recorded Canva Pro rising from $144 to $180 a year in early August 2026. It then recorded a return to $144 in mid September. The same September snapshot recorded the Free plan cut from up to 200 Standard or 20 Premium uses to up to 20 uses of either.

The September change also moved Standard AI off the meter on Pro, Business and Enterprise, under a fair-use clause. In plain terms, Canva cut the price back, made basic AI unmetered, and kept the expensive generation metered. That pattern of quietly reshaping allowances rather than headline price is what the analysis of AI shrinkflation and usage limits calls the second price.

Five design tasks priced on each AI design tool

Here is the comparison the credit labels hide. We chose five tasks a design team does weekly and priced each one from the vendors' published rates. Where a vendor has no equivalent feature or does not publish the rate, the cell says so rather than guessing.

Figma figures are from its rate table dated 6 October 2026; "+" marks Figma's own approximate floors. Adobe partner-model rates are from its partner-model help page and may include limited-time pricing. Canva tier assignments are from its help centre as indexed by search. Units are not interchangeable across columns.

For still images, Adobe wins on the house model. If your team's image work is mostly mood boards, placeholders and textures, Firefly's unlimited standard model makes the marginal cost zero. Figma's cheapest image model is close behind at 2 credits, which gives a Professional seat up to 1,500 images a month before anything else is spent. Canva's Premium image use is the most rationed of the three on paper.

Agentic layout work is where Figma meters hardest. The 300+ credit floor for a five-screen flow means a Professional seat gets at most 10 such runs a month, and fewer if the designer also generates images. Agentic features only joined the meter on 6 October, so nobody has a full month of real usage data yet.

What an entry paid seat buys in a month

A per-task cost becomes a budget once you divide it into the monthly allowance. The table below does that for each vendor's entry paid tier, using the cheapest published rate for each task.

"At most" because Figma's agentic costs are floors. Canva counts are third-party figures and come from one shared pool, so Premium and Ultra numbers are alternatives, not a sum.

Converting credits to dollars is possible for one vendor only. At Adobe's $9.99 add-on rate, a Nano Banana Pro image at 40 credits costs about $0.20, a GPT Image generation at 60 credits about $0.30, and a five-second clip at roughly 100 credits about $0.50.

Figma and Canva bundle AI into a seat that also buys the core design product. As an illustration only: if you assigned the entire Canva Pro subscription of $144 a year to its 20 monthly Ultra uses, each would carry a ceiling of $0.60. That number overstates the true cost, because nobody buys Canva only for Ultra. The wider argument for why bundled AI hides its unit price is laid out in the comparison of AI bundling versus metering.

Where this comparison is weakest

Here are the specific places the evidence is thin.

The Canva numbers are one step removed

Canva's pricing page and help centre both refused automated access during our research. The tier assignments come from Canva's indexed help-centre text. The per-plan counts of 200 Premium and 20 Ultra uses come from third-party sites that agree with each other, not from a Canva page we opened. If you are budgeting Canva, open the AI usage article in your own account before relying on our column.

Seats buy more than AI, and promotions distort the rest

A Figma Professional seat buys the whole design tool, and the AI allowance is a component of it. Comparing it to an Adobe Firefly plan, which is mostly an AI product, flatters Adobe on a per-credit view. Adobe is also running first-year promotions at the time of writing, including unlimited generations on select partner models for new Pro subscribers. A team that signs today may see costs jump at month 13. Our five tasks were chosen by us, and a team that mostly edits photos would weight them very differently.

Which meter suits which design team

The right tool is the one whose free zone covers most of your volume.

If your work is high-volume still imagery, Adobe's unlimited house model is the safest budget. We would choose it over the other two on cost predictability alone. If your team prototypes product interfaces, Figma's per-seat meter is the one to model carefully. Its agentic features are the newest and the most expensive, and its overage price is hidden until you are inside the admin dashboard.

If your team is marketing-led and mixes templates with occasional generation, Canva's move of Standard AI onto fair use makes it cheap for routine work. Its Ultra tier is the constraint. Twenty video clips or premium images a month disappears quickly in a campaign week. The HubSpot Breeze credits repricing did the same to support teams: the price fell, and the variance rose.

One structural point cuts across all three. Every vendor is shifting from charging per person to charging per action, the same move that the analysis of seat compression in SaaS pricing traced across business software.

Frequently asked questions

How to budget AI design credits this quarter

Start with an export, not a vendor call. Pull last month's AI usage by person from whichever tool your team uses most. Figma's admin dashboard holds usage history back to November 2025, which gives you a baseline before the 6 October agent metering began.

Then sort that usage by task, using the five rows in the table above. If most of your volume is still images, ask whether a cheaper model or Adobe's house model covers it. If most of it is agentic layout work, set a pay-as-you-go spending limit before the first designer hits the seat cap in the middle of a sprint. Forecasting that kind of variable AI spend is the subject of the review of AI spend forecasting tools.

Related reading