AI limits glossary
Plain-language definitions for the terms behind AI usage limits.
- Token
- The unit AI models use to measure text — roughly ¾ of a word. Usage limits and costs are counted in tokens, so shorter prompts and replies stretch your limits further.
- Rate limit
- A cap on how much you can use a model in a given window. Hitting it pauses your access until the window resets.
- 5-hour session limit
- A rolling usage window (used by Claude) that resets five hours after it starts. LimitWise shows how much of it you have left and when it resets.
- Weekly limit
- A longer usage cap measured over seven days, often applied across all of a provider’s models. LimitWise tracks it alongside shorter windows.
- Context window
- The maximum amount of text a model can consider at once. When a conversation gets long, older messages can fall out of the window.
- Depletion velocity
- How fast you are using up a limit. LimitWise estimates it so you can pace a long session instead of getting locked out unexpectedly.
- Cross-LLM handoff
- Moving an active conversation from one AI model to another. LimitWise’s Omni-Bridge does this in one click, any model to any model.
- Prompt optimization
- Rewriting a prompt to be shorter and clearer without changing its intent, so it uses fewer tokens. LimitWise does this on-device.
- Local-first
- A design where data and processing stay on your device instead of a remote server. LimitWise is local-first: your context and usage data never leave the browser.
- Multi-account tracking
- Monitoring usage for several independent AI accounts at once — useful when you run more than one subscription.