Body
Questions
- How do I reduce my token usage in Claude?
- Why am I hitting my usage limit so quickly?
- How can I make Claude cheaper to use?
- What are tokens, and why do they matter?
- Does a long conversation cost more than a short one?
- How should I attach files or documents without wasting tokens?
- What are some tips to get good answers while using fewer tokens?
- How do I keep Claude usage costs down for my research or department?
Environment
This article applies to Bowdoin community members using Claude who want to work efficiently and stay within their usage limits or budget.
- Who: Students, faculty, and staff using the Claude apps, Claude Code, or Cowork.
- What: Practical habits that reduce token use. These habits matter most when your usage is measured against a grant, a department budget, or an application programming interface (API) allocation.
Resolution
A token is a small chunk of text, roughly three-quarters of a word. Claude measures usage in tokens, counting both what you send (your prompt, any attached files, and the whole conversation so far) and what Claude generates in reply. Because Claude re-reads the entire conversation on every turn, long chats and large attachments add up quickly. The habits below reduce token use without hurting quality.
Manage the conversation
- Start a new chat for a new topic. A long thread carries all previous messages forward and re-processes them on every turn, so a fresh topic deserves a fresh chat.
- Keep threads focused. Finish one task, then start a new conversation rather than piling unrelated questions into a single long thread.
- Summarize instead of continuing forever. If a chat grows long but you still need its context, ask Claude to summarize the key points, then paste that summary into a new chat.
Write efficient prompts
- Be specific up front. Clear instructions get the right answer on the first try and avoid several rounds of corrections, which is where token use quietly piles up.
- Ask for the length you need. Request "a short summary" or "just the code, with no explanation" when that is all you want.
- Batch related questions into one well-structured message instead of many small back-and-forth turns.
Handle files and context wisely
- Attach only what is needed. Paste or upload the specific section you need rather than an entire book or code repository.
- Re-use, do not re-paste. Once a document is in the conversation, refer to the document rather than pasting it again.
- Prefer targeted excerpts over large files when you only need a few pages.
Choose the right model
Note: Model choice is the biggest single lever on token use. Lighter models use far fewer tokens. Use Haiku for simple tasks, Sonnet for everyday work, and reserve Opus for problems that truly need deep reasoning. For details, see Choose the Right Claude Model in the Related Articles section.
If your usage is funded or metered
Important: If your usage is funded by a grant or a department budget, token use draws against that funding source. Apply the habits above, prefer lighter models for routine work, and monitor your balance. For API or Portkey usage, check your allocation before running large or repeated jobs.
Additional Help
If you need further assistance, you have several options:
- Bowdoin Bot: Chat with Bowdoin Bot directly from any KB page for instant answers.
- Phone: Call the Bowdoin College Service Desk at (207) 725-3030.
- In person: Visit the Tech Hub in Smith Union during business hours.
- Submit a ticket: Request assistance through the Service Catalog.
AI-assisted content: This article was drafted with the assistance of an AI writing tool and reviewed by Bowdoin IT staff for accuracy.