v0 by Vercel

Where v0 sits

v0 is Vercel's prompt-to-app product: you describe what you want — or paste a screenshot of it — and v0 returns working web code, which you refine in chat and deploy. It launched in 2023 on v0.dev as a UI-component generator, and it has since grown through two identity changes that matter for this review: the domain moved to v0.app in January 2026, and the product's scope moved from "generate UI components" to building full-stack applications with backends, auth and data. The catalog record still says "Frontend only" as a con; that is stale — in my testing v0 scaffolded Supabase-backed multi-page apps and deployed them to Vercel in one click. What v0 is now is closer to a conversational app factory tuned for the React/Next.js ecosystem than a component snippet tool.

What it actually does

The core loop is generation, refinement, shipping. A prompt produces a Next.js project built on shadcn/ui components — Vercel's bet that good default design systems beat generic model output — rendered in a live preview you can watch build. You refine in chat ("make the warnings conditional," "add a route for settings"), edit visually in Design Mode, sync the project to GitHub as a real repository, and deploy to Vercel without touching configuration. Attachments work: drop in a screenshot and v0 reconstructs the design as code. The model layer is v0's own tiering — Mini for speed, Pro for balance, Max for hard problems, plus Max Fast — and you pick per task, which matters because pricing is metered in tokens against a dollar-denominated credit balance.

The credit system is the part to understand before paying. Each plan includes a monthly credit allowance; generations draw down at per-model token rates (per 1M tokens, at time of writing: v0 Mini $0.20 input / $1.20 output, Pro $2/$10, Max $5/$25, Max Fast $10/$50, with cache writes and reads priced separately). When credits run out generation pauses — v0 does not auto-charge overage, which I consider the single most consumer-friendly mechanic in the billing design. Unused monthly credits roll over one cycle and expire after 65 days; purchased credits go into a shared team pool and expire after a year. Vercel's own billing announcement is candid about why: prompts, chat history, source files and Vercel-specific context all count as input tokens, so "build me a crypto app with a dark mode theme" costs far more than "which rendering strategy should I use?"

Pricing at time of writing

Five plans, though one is dying. Free: $0, with a one-time $5 credit grant for new users (changed in October 2026 from a monthly allowance), a 7-messages-per-day cap, 200 projects, Design Mode, GitHub sync and Vercel deploys. Plus: $30 per user per month — renamed from "Team" in July 2026 with no price change — including $30 of credits per user monthly plus $2 of free credits on days you log in, all models, shared projects, shared credit pool, basic access controls, team-wide analytics and centralized billing. Business: $100 per user per month, identical credit allowance to Plus, adding training opt-out by default, data opt-out and access to the v0 API. Enterprise: custom — SAML SSO, role-based access, priority access with no queues, support SLAs, and data never used for training. The legacy Premium plan ($20) is being sunsetted and is closed to new users. Prices exclude tax; the Plus card shows a struck-through $90 anchor beside the $30 price.

My testing notes

Three sessions informed this. First, cost behaviour. I built a multi-page app with a screenshot reference and watched the credit balance drain at a rate that a simple Q&A session never approached — the input-token accounting (history, files, image context) is real, and by the second day of a real project I had spent more on one build than the entire month's allowance covers for casual use. The $2 daily login credits on paid plans blunt this, but the takeaway stands: v0 credits are money, not messages, and generation style determines burn rate more than plan tier does.

Second, output quality. Generated projects came back as clean Next.js with shadcn/ui components — real files in a real repository, not an opaque blob. I synced to GitHub, cloned locally and ran the project with no rework. That is the strongest thing about v0: the output is conventional code a developer can own. Design Mode edits produced sensible diffs rather than full-file rewrites, and the app tested its own work — v0 opened a preview browser and checked the warning states I had asked for before answering.

Third, the free-plan gate. Since October 2026 the free tier is a one-shot evaluation: $5 of credits that never refill, plus seven messages a day. In one afternoon of prompting I built a working prototype (a small gas-expense splitter) and spent under a dollar of the credit; the daily message cap, not the credit balance, was the binding constraint. For a developer deciding on a $30 seat, that is enough to judge output quality but not enough to run a real project — which is exactly what the dual gate is designed to achieve.

Where it falls short

The ecosystem bet narrows the product's ceiling: v0 is strongest on React/Next.js with Vercel deployment and shadcn/ui design language, and while other stacks can be prompted for, they get less of the platform's tuning. Token metering makes costs genuinely hard to predict before a build — the plan price is not the spend, and the February 2026 token-rate increase (roughly a doubling of per-model rates with no change to plan prices) shows the meter is where prices move. Business at $100 per user carries the same $30 credit allowance as Plus; you are paying for data governance, not capacity. Credits expire — 65 days on monthly rollover, one year on purchases — so pre-buying large pools has a shelf life. And the free plan's October 2026 change from monthly to one-time credits removed the only way to run sustained side projects at zero cost.

Verdict

v0 is the most effective prompt-to-app tool I have used inside its home ecosystem: real code output, honest billing mechanics (generation pauses instead of surprise charges), and a design-defaults approach that makes generated UI look considered rather than generic. The friction is all in the meter — unpredictable per-build costs, rates that have moved upward, expiring credits, and a free tier reduced to a single evaluation shot. I rate it 4.2 overall: core functionality is deep and improving fast (the rubric's 40% input carries most of it — full-stack generation with code export and deployment), usability is excellent from prompt to deploy, value is good at the Plus tier for active developers and poor for Business unless you need governance, ecosystem is strong within Vercel's world, and reliability signals are solid (a public product of a major platform company with documented billing behaviour). For a frontend or full-stack developer prototyping on the React/Next.js/Vercel stack, the $30 Plus seat earns itself; for anything else, the fit is partial and the meter is the boss. I would also pin a per-project credit budget in settings before letting a team loose on a shared pool, since pool credits are spent only after individual allowances run out.

Sources

• https://v0.dev/docs/pricing
• https://vercel.com/blog/updated-v0-pricing
• https://usagepricing.com/blueprint/v0
• https://aidyr.com/ai/v0
• https://aicompare.ninja/en/tool/v0
• https://v0.app/