Leonardo AI

What Leonardo actually meters

Leonardo meters a token whose price moves with computational intensity. The help centre says each feature and model carries its own cost, set by complexity and computing power and varying with reference images, settings, image size and batch size. The published table covers the older image pipeline:

Reference images, upscaling, batch size and the premium pipelines drive the number up, and batches aren't linear: four Alchemy images cost 26, not 64.

Here's the gap. That table stops at features Leonardo shipped years ago. For the models people use now, Veo 3.1, Kling, Seedream, Nano Banana, Wan, no token cost appears in the help centre, the docs or the pricing page. Leonardo prints the cost on the generate button instead: honest when you spend, useless when you budget.

How the token bank works

Unused tokens bank, up to three months of allowance. Whatever's left at the end of a cycle moves into the Token Rollover Bank, capped at 25,500 on Essential, 75,000 on Premium and 180,000 on Ultimate, exactly three times each monthly grant. Teams bank per seat on the same multiple.

Free accounts get none of it. The daily 150 tokens expire every 24 hours, and the pricing page's "Token Bank 150" row for Free is simply wrong. Leonardo's Token Rollover article, updated 28 July 2026, says the Bank only stores tokens on paid plans with an active subscription.

Top-up packs never expire and ignore the rollover cap, but you can't buy or spend them without an active subscription, and Teams can't buy them at all. Pack prices aren't published.

Deduction order is published and unusual: Rollover Bank, then Fast tokens, then top-ups. Draining the bank first protects the pool that can expire, which is the customer-friendly ordering.

Cancelling cancels the bank. Downgrading expires anything above the lower cap. Upgrading keeps the balance and raises the ceiling. Pausing freezes it and drops you to 150 daily tokens.

What happens when you hit the limit

Generation stops, and what follows depends on the tier. Premium falls back to unlimited Relaxed image generation and Ultimate adds Relaxed video, both slowing under load. You can't switch Relaxed on by hand; it activates only once every token is gone.

Relaxed excludes third-party models, and the list is long: GPT-Image-1 and 1.5, Flux.1 Kontext and Kontext Max, FLUX.2 Pro, Ideogram, Nano Banana, Kling, Seedream, Sora 2 and Veo 3.1. Those always need tokens.

Essential and Free have no fallback. A paid account with an empty bank and no top-ups drops to the same 150 daily tokens a free user gets. Nothing auto-charges. On the API, a zero balance pauses requests unless auto top-up is set.