Best Practices for Messaging Input Tokens

[vijarkohli-goldendoo (Vijar)]

Are there any best practices for messaging? With the new pricing system, I’d like to maximize my input tokens vs output tokens.

For example, I’m planning to lock specific files and folders so v0 doesn’t execute too many tasks. I’m opening to leveraging other tools like Cursor to help reduce my v0 variable costs. Feel free to share other programming tools or LLM tips.

P.S. ELI5 I’m not an expert developer so there are several layers of programming I’m not familiar with yet.

[pawlean (Pauline P. Narvas)]

Hi, Vijar!

Great discussion point! Thanks for kicking it off.

I’ll let folks jump in as well, but this what I’ve found has worked for me:

1. Be specific in your requests - Instead of “make me a website,” try “create a simple landing page with a hero section and contact form using Next.js and Tailwind”
2. Break down complex tasks - Rather than asking for an entire application at once, build it piece by piece in separate conversations
3. Reference existing code - Point v0 to specific files you want modified instead of asking it to generate everything from scratch

Some additional resources:

• Vercel v0 Documentation
• Vercel v0 Pricing
• Become a v0 Expert

[pawlean (Pauline P. Narvas)]

I also wanted to follow up with more tips after we shared this with the v0 team:

You can set a custom instruction in v0 to output only code when building applications. This helps reduce token usage by skipping explanations about code changes if those aren’t needed.

We’re also introducing v0 small soon—a lighter-weight option for simpler tasks that don’t require high-level intelligence, making those tasks more cost-effective.

Additionally, smaller projects and modular file structures tend to use fewer tokens, especially during edits, compared to working with large, monolithic files like a single page.tsx.

Hopefully that helps!

[vijarkohli-goldendoo (Vijar)]

Thanks, Pauline. The custom instructions are a good idea. I haven’t used them enough.

It would be helpful if we had details into the token count per message. For example, adding a new page might be ~50-100k tokens and $0.30-0.35 to complete. You can add a token estimate counter within the chat sidebar.

I look forward to testing the lighter-weight option soon.

Btw, I’ve never found the enhance button to be useful. It creates very long prompts and v0 seems to prefer short, clear instructions.

[pawlean (Pauline P. Narvas)]

We shipped this btw.