v0

What Is v0?

v0 is Vercel's AI builder. You describe an interface or an application, and it writes the React code — not a picture of a component, the component itself, in the same stack a Vercel-hosted product would use anyway. You can then edit it visually, push it to GitHub, and deploy it without leaving the page.

That last part is the difference between v0 and the general-purpose app builders it competes with. It is made by the company that runs the hosting, so the path from prompt to a live URL has no seams in it. The price of that smoothness is an opinion about your stack, which you are adopting whether or not you meant to.

It began as a component generator and grew into something that will build a full application with data and server logic. The component work is still where it is most reliably excellent.

How It Works

Prompt to Real Code

Describe a pricing table, a navigation bar, a login form, a dashboard — you get working React with sensible structure and modern styling conventions rather than markup that only looks right in a screenshot. Iterate by describing the change, or fork the result and take it in a different direction.

Because the output is ordinary code in a conventional stack, a developer can read it, correct it and keep it. This is the practical dividing line between AI builders worth using and AI builders that generate something nobody can maintain.

Design Mode

Once something is generated, you can adjust it directly — spacing, text, colours — instead of writing another paragraph of prompt hoping the model understands «a bit tighter». For the last ten per cent of a layout, which is where prompting is worst, this is the feature that saves the most time and the most credits.

GitHub and Deployment

Projects sync to GitHub and deploy to Vercel, including on the free tier. The escape hatch is genuine: whatever you build here is a repository you can clone and continue in your own editor, which makes v0 usable as a starting point rather than a place you have to stay.

Four Models, Four Prices

v0 offers its own model line-up rather than one setting: a small fast one for routine work, a balanced default, a maximum-capability tier for complex tasks, and the same maximum capability at higher speed. They differ by roughly fifty times in output cost between cheapest and dearest, so choosing deliberately matters more here than in most tools.

How Billing Actually Works

This is the part that surprises people. v0 does not bill by message or by project — the subscription buys a wallet of credits, and every message spends from it at the token rate of whichever model you chose.

Cached input is far cheaper than fresh input, which rewards staying in one conversation rather than starting over. And the reason short prompts consume large numbers of input tokens is that each message carries your project's context, not just the sentence you typed — a five-word instruction on a large project is not a five-word request.

Run out of included credits and you buy more on demand, shared across the team. The consequence worth planning for: two people on the same plan can spend wildly different amounts depending on which model they leave selected.

What It Costs

Look closely at Plus: thirty dollars a month buys thirty dollars of credits. You are not paying a margin on usage — you are paying for the account, the team features and a predictable ceiling, and the credits are effectively the whole fee. That is unusually honest pricing.

Business is the row to think hardest about. It costs more than three times Plus and includes the same thirty dollars of credits. The material difference is that your work is opted out of training by default. If that matters to your organisation it is worth the money; if it does not, you are paying seventy dollars a user for nothing you will notice.

The free tier is a real trial — five dollars of credits, deployment included — but seven messages a day is a hard stop rather than a soft one. It answers the question «is this any good» and cannot answer «can I finish this today».

Who Gets the Most From It

Frontend Developers

The clearest case, and not the one people expect. v0 is at its best generating the components you would otherwise write from memory — forms, tables, modals, empty states — in code you can actually review. It removes typing, not judgement.

Teams Already on Vercel

If your product already runs on this stack, v0 has no integration cost at all. The generated code belongs in your repository as it is, and the deployment is the one you already use.

Designers Who Want Working Interfaces

A real, responsive, interactive page instead of a static frame — and Design Mode means you can adjust it yourself without asking anyone to rebuild it. For validating an interface with actual users, this is a different class of artefact.

Anyone Prototyping Under Time Pressure

Client pitch tomorrow, internal review this afternoon. Describe it, deploy it, send the link. The fact that the link is a real application rather than a mockup changes how the meeting goes.

What to Watch Out For

• It has strong opinions about your stack. If you are not building in React and deploying to Vercel, much of the advantage disappears.
• Costs are token-based and vary about fiftyfold between the cheapest and dearest model. Leaving the top model selected is an expensive habit.
• Short prompts are not cheap prompts — your project's context is sent with every message.
• Business costs over three times Plus for identical credits. The difference is training opt-out, nothing else material.
• The free tier's seven-message daily cap stops work rather than slowing it.
• Generated code is good but not automatically accessible, secure or performant. Review it like you would a junior developer's pull request.
• Full-stack generation is less reliable than component generation. The further from UI you go, the more you will be correcting.
• Credits are shared across a team, so one person exploring on the expensive model can drain the month.

Frequently Asked Questions

Is v0 free?

There is a free plan with five dollars of monthly credits, Vercel deployment, Design Mode and GitHub sync, limited to seven messages a day. It is a genuine trial rather than a workable long-term plan.

Do I get the code?

Yes. Projects sync to GitHub, so the output is a repository you can clone and continue anywhere. That is the main reason to prefer it over builders whose output cannot leave the platform.

What does it cost per message?

It depends on the model and on how much context your project carries. Output tokens range from about a dollar twenty per million on the small model to fifty on the fastest top-tier one, so the same request can differ in cost by more than an order of magnitude.

How is it different from Lovable or Bolt?

v0 is narrower and deeper: it is made by the hosting company, targets one stack, and is strongest at interface work. The general builders will do more kinds of application; v0 will produce React you are happier to keep.

Is my work used for training?

On Business, training opt-out is the default. On Enterprise, data is never used for training. On the lower plans, check the current settings and terms rather than assuming.

Do I need a Vercel account?

Billing runs through vercel.com and deployment goes there by default, so in practice yes. You can still take the repository elsewhere and host it wherever you like.

The Bottom Line

v0 is the AI builder for people who were going to write React anyway. It generates interfaces good enough to keep, lets you fix the last ten per cent by hand instead of by prompt, and puts the result in GitHub where it belongs. Within its stack it is the strongest tool of its kind; outside it, most of the argument evaporates.

Watch the model selector more carefully than the plan selector — that is where the money goes. Start on Plus, keep the cheaper model for routine work, and only pay for Business if training opt-out is a requirement rather than a preference.