AI Usage and Limits

What are usage limits?

Revise offers AI models from OpenAI (GPT), Anthropic (Claude), Google (Gemini), and xAI (Grok). These models all cost different amounts of money for every word they process, and they have other characteristics that make them behave differently from each other in terms of cost.

Accordingly, paying Revise customers get allocated a monthly limit which they can spend using whichever models they like. This usage is measured internally in the actual dollar cost of your requests to the AI providers used by Revise.

You can see how much of your monthly limit you have left in your Account dropdown in the top right corner of the screen:

You can view your current AI Usage in your account dropdown

If you are using Revise on the free tier, rather than a monthly limit you have a much smaller daily limit which resets every 24 hours. Please updgrade to Plus or Pro to use Revise in longer sessions with only a monthly limit.

How is the AI limit tracked?

Usage is based on the real cost of the work the AI agent does for you, not the number of messages you send. A small edit with a fast model will use very little of your limit compared to a large rewrite of a longer document. Using more expensive models will also use your limit faster.

How much do different models cost?

This table shows approximately how much of your limit each model will use compared to GPT 5.4 Mini, which is the default model for Revise. Note that these are based off internal benchmarks of real AI agent work - not just API prices from the providers! Some models are more efficient than others, so they cannot be compared purely on per-token costs.

While these numbers are based on real benchmarks, they are just an approximation. Actual usage will still vary with prompt length, document context, attachments, web search use, and how complicated your requests are.

Model Multiplier Breakdown

What uses more of your limit

• Always using a high-end model - for example, setting your model to Claude Opus and never changing it will use your limit much faster
• Long documents and large selections - more text gives the model more to read.
• Large attachments - PDFs, images, Word files, and big pasted references add context the model needs to process.
• Open-ended requests - broad prompts like "improve this whole document" can require many reads, edits, and checks.
• Web search and document search - retrieving and reading outside context adds work to the turn.
• Extended Mode - useful for bigger jobs, but it gives the agent more room to plan, call tools, revise, and verify.

How to use your limit efficiently

• Start with a fast, low-multiplier model for routine edits, summaries, formatting changes, outlines, and brainstorming.
• Switch to a stronger model when the task needs deeper reasoning, careful synthesis, or a final quality pass.
• Select the exact passage you want edited instead of asking the agent to scan the whole document.
• Ask for one clear outcome at a time: "shorten this section by 20%" is usually more efficient than "make this better."
• Use headings and sections in your document so the agent can find the right place quickly.
• Attach only the pages, excerpts, or files that matter for the current request.
• Use the AI Rewriter for quick standalone rewrites of shorter text snippets. AI Rewriter is simpler and uses much less of your monthly limit than the main document editor interface.

When to use each model type

In general, cheaper models are good for simpler edits while more expensive models are best for changes that require more thought, research, or lots of creative writing.

If you reach your limit

If you reach your monthly limit and want to keep using Revise, you can upgrade to Revise Pro for a higher limit. If you're already using Revise Pro, you can enable Extra Usage in Billing Settings to keep using AI on a pay-as-you-go basis for the rest of the month.

Free tier usage resets on a rolling daily window - so if you don't want to upgrade yet you can just come back tomorrow!

If you are close to the limit, use less expensive models and keep your prompts focused.