If you’re on a Runway annual Unlimited plan, stop reading reviews and go check your account — you have a decision to make this month, and the window closes at the end of August 2026.
That’s an unusual way to open a Runway ML review, but it’s the most useful thing anyone can tell you right now. Everything else about this platform — the model quality, the editor, the credit arithmetic — can wait a week. The billing change can’t.
Unlimited is being retired. Max replaces it.
According to pricing trackers checking Runway’s own transition documentation in August 2026, the sequence is:
- Existing Unlimited plans run unchanged through 31 August 2026.
- Monthly subscribers move to Max on 1 September, at the same price.
- Annual subscribers choose during August: a refund covering the remaining months, or a credits bonus with migration to Max.
Max is not a rename. It’s a different product:
| Unlimited (retiring) | Max (replacement) | |
|---|---|---|
| Monthly credits | 2,250 | 9,500 |
| Explore Mode | Yes — unlimited generations at a relaxed queue rate | No |
| Rollover | None | Up to one month of unused credits |
| Ceiling | Soft — slow, but you could keep going | Hard — 9,500 credits and you stop |
Runway’s stated reason for the change relates to queueing. Whether four times the credits compensates for losing an uncapped lane depends entirely on how you work. If you iterate constantly — generating twenty variations to keep one — Explore Mode was the whole point of the plan, and 9,500 credits is a fixed number that runs out. If you generate deliberately and rarely queued anything overnight, Max is straightforwardly more.
Run the arithmetic on your own last three months before you choose. That’s a better basis than any recommendation, including this one.
Why Unlimited was always a strange product
Worth understanding, because it explains the retirement.
Users noticed something on Runway’s own pricing page: Unlimited at $95 a month included 2,250 credits — exactly the same allowance as Pro at $35. The entire $60 difference bought Explore Mode and nothing else. Community discussion at the time also reported accounts being suspended for heavy use under a plan called Unlimited, which is the sort of thing that ends a pricing experiment.
So the honest read is that Runway sold a queue lane and called it unlimited, discovered what happens when you do that, and has now converted it into a large but finite credit bucket. That’s a more defensible product. It’s also a worse deal for the specific users who bought the old one on purpose.
The credit arithmetic nobody prints
Everything on Runway runs on credits, and the rate depends on which model you use. Rates reported from Runway’s plan equivalency data in May 2026:
| Model | Credits per second | 10 seconds costs |
|---|---|---|
| Gen-4.5 (flagship) | 25 | 250 credits |
| Gen-4 | 12 | 120 credits |
| Gen-4 Turbo | 5 | 50 credits |
Now put that against the entry tier. Standard ships roughly 625 credits a month. At the flagship rate that is about twenty-five seconds of video per month. Not twenty-five clips — twenty-five seconds, total.
And two multipliers make it worse than that sounds. Failed generations still spend credits. A prompt that produces something unusable costs the same as one that works, and in generative video the hit rate is not high. Plan on keeping perhaps one clip in four or five, which turns twenty-five seconds of allowance into five usable seconds.
Credits don’t roll over on any tier except Max, which carries up to one month.
The practical technique that follows: draft on Turbo, finish on the flagship. At 5 credits a second versus 25, you can iterate five times as much while you’re figuring out composition and motion, then spend the expensive model only on the take you’ve already chosen. Doing this is the difference between Standard being usable and Standard being a demo.
Why every Runway pricing page disagrees with every other one
This is worth stating plainly, because you will notice it the moment you compare two sources.
Runway renamed its tiers during 2026 — reporting suggests the old Basic became Standard, old Standard became Pro, and old Pro became Max, with the top allowance raised. It also adjusted plan inclusions at least twice in six months and changed at least one model’s credit rate. The result is that trackers published in March, May and August 2026 describe genuinely different products using overlapping names, and all of them look current.
You’ll find Standard quoted at $12, $15 and $28. Pro at $35 and $76. A $188 Unlimited tier that reportedly no longer exists. At least one tracker states Gen-4.5 dropped from 25 to 12 credits per second, which others don’t reflect.
I can’t resolve that from outside, and neither can any other review — so treat every number on this page, including mine, as a shape rather than a quote. What’s stable is the structure: a small free allowance that doesn’t refresh, three or four paid tiers separated by credit volume, per-second rates that differ by model, and no rollover below the top tier. Confirm the actual figures on Runway’s own pricing page at checkout. This is the same discipline that cutting business costs with AI depends on: the billing model is the product decision.
What you’re actually buying in 2026
Here’s the reframe that matters more than any tier comparison.
Runway’s flagship model is no longer the reason to buy Runway. The 2026 video field is crowded — Sora 2, Veo 3.1, Kling 3.0 and others all shipped or upgraded during the year, and on raw generation quality the flagship spot moves every few months.
What Runway did in response was become a storefront as well as a studio. Paid plans bundle access to third-party models — Veo, Kling, Seedance among them — all drawing from the same credit pool, alongside Runway’s own Gen-4 family, Act-Two, Aleph and the editing environment.
That changes the buying question from “is Gen-4.5 the best model” to “is one subscription and one meter across several models worth more than the cheapest single model.” For an agency juggling client work, usually yes. For someone who wants one specific model, usually no — the standalone tools are cheaper.
It also means the editor is doing more of the work than the marketing suggests. Motion brush, director controls, frame interpolation and the timeline are what production people actually cite as the reason to stay, and they’re the part competitors haven’t replicated. If your workflow is cutting existing footage rather than generating it, that’s a different tool entirely — see Descript vs CapCut.
The free plan is a demo
Runway’s free tier gives 125 credits, one time only. They don’t refresh monthly. Output carries a watermark, the model selection is limited, there’s no commercial licence, and free accounts reportedly can’t buy top-up credits.
125 credits is roughly five seconds of flagship output, or around 25 seconds on Turbo. It’s enough to see whether the interface suits you and nothing more. Budget from Standard upward from the start.
Who should buy which
| If you are | Tier | Reasoning |
|---|---|---|
| Testing whether AI video fits your work at all | Free, then Standard | Draft everything on Turbo or the allowance evaporates |
| A creator producing a few short clips a month | Standard | Workable only with disciplined model selection |
| Producing regularly for clients | Pro | Where the credit ceiling stops interrupting the work |
| Iterating heavily every day | Max | The biggest bucket, with one month of rollover — but a hard ceiling now |
| A team needing SSO and shared workspaces | Enterprise | Separate accounts save money but cost you shared libraries and billing |
| Making talking-head or narrated content | None of these | Wrong tool — see below |
That last row matters. Generative video is for shots, not presenters. If you need a person delivering a script, Synthesia is the category, and if you need the voice alone, ElevenLabs vs Murf covers the meters there.
Honest limitations
Cost per usable second is the real number. Not cost per second generated. Multiply the published rate by your actual keep rate, and be pessimistic about it.
Consistency across shots remains hard. Characters, lighting and props drift between generations. This is the structural reason AI video is used for inserts and B-roll rather than for narrative sequences.
The credit pool is shared across everything. Video, image, upscaling and third-party models all draw from one allowance, so an afternoon of image experiments quietly costs you video.
Commercial and IP posture is your own problem. Training-data litigation across the generative-media sector is unresolved, and the ownership question for purely AI-generated output is separate from the infringement question. Both are laid out in Midjourney vs DALL·E vs Firefly, and the reasoning transfers directly to video.
Pricing has moved twice in six months. Whatever you buy, don’t assume the terms are stable for a year.
The verdict
Runway is the most complete AI video environment available, and increasingly the value is the environment rather than the model — an editor professionals prefer, plus one meter across several competing models. That’s a real proposition for agencies and production teams.
For individual creators it’s harder to justify. The entry tier buys seconds, not minutes, and the discipline required to make it work — draft cheap, finish expensive, accept a low keep rate — is real skill rather than a setting. If your output is social video, the honest starting point is making AI videos without editing skills, and the economics of what actually pays are in AI tools for content creators. For the wider stack, the best AI tools for small business owners is the map.
And if you’re on annual Unlimited: go and make that choice before the month ends.
Frequently asked questions
Is Runway’s Unlimited plan gone?
It’s being retired. Existing Unlimited plans reportedly run unchanged through 31 August 2026; monthly subscribers move to Max on 1 September at the same price, and annual subscribers choose during August between a refund for the remaining months or a credits bonus with migration. Max carries far more credits — around 9,500 versus 2,250 — but does not include Explore Mode.
What was Explore Mode and does Max have it?
Explore Mode allowed unlimited generations at a relaxed, lower-priority queue rate, and it was the only meaningful difference between Unlimited and the cheaper Pro tier, which carried identical credits. Max does not include it. Every generation on Max draws from the credit allowance.
How much video do Runway credits actually buy?
At reported rates of 25 credits per second for the flagship model, 12 for Gen-4 and 5 for Gen-4 Turbo, an entry tier of around 625 credits is roughly 25 seconds of flagship output a month. Failed generations spend credits too, so plan on considerably fewer usable seconds than the ceiling implies.
Why do Runway pricing guides contradict each other?
Because the tiers were renamed during 2026 and plan inclusions changed at least twice in six months. Guides written months apart describe different products under overlapping names, and at least one model’s credit rate is reported differently across sources. Verify the current figures on Runway’s own pricing page rather than trusting any third-party summary.
Is the free plan enough to evaluate Runway?
Barely. It provides 125 credits once rather than monthly, output is watermarked, model access is limited, there is no commercial licence, and free accounts reportedly cannot buy additional credits. That’s roughly five seconds of flagship output — enough to judge the interface, not the results.
Should I use Runway or a cheaper single-model tool?
If you want one specific model, standalone tools are cheaper. Runway’s 2026 proposition is the editing environment plus access to several competing models — including third-party ones — on a single subscription and a single meter. That bundling is worth paying for when you produce regularly across varied briefs, and not otherwise.
