AI tools like Claude, Dartmouth Chat, Microsoft Copilot, ChatGPT, and Gemini all measure usage in tokens rather than words or messages. This article explains what tokens are, why they matter, and how to get the most out of your daily usage across these tools.
What Is a Token?
A token is a small chunk of text, roughly 4 characters or about three-quarters of a word. Short, common words like "the" or "is" are usually one token each. Longer or less common words may be split into two or more tokens. Uploaded files, such as PDFs and images, use tokens too, since the AI has to "read" that content before responding.
Why Tokens Matter
Tokens matter for two reasons: cost and memory. Most AI tools are priced based on token usage, and institutional licenses (like the ones Dartmouth provides) often share a token or credit budget across the community. Tokens also determine the context window, which is how much of your conversation the AI can "remember" at once, including everything you have typed and everything it has replied with. Long conversations, large uploads, and lengthy AI responses can fill that window faster than expected, causing the AI to lose track of earlier instructions or documents.
How Token Limits Work Across Dartmouth's AI Tools
- Claude for Education (claude.dartmouth.edu): Token limits refresh every 5 hours. Simple questions use very few tokens, while large uploads, long conversations, or multi-step agentic tasks can use significantly more. Using less intensive models such as Sonnet or Haiku rather than Opus or Fable 5 when appropriate will stretch those tokens for a longer time.
- Dartmouth Chat (chat.dartmouth.edu): Usage is measured in credits, with a daily limit of 2,000 credits that resets every 24 hours. Local models (such as Mistral and Llama) have no usage limit; cloud models (including Claude, GPT, and Gemini models) draw from the daily credit limit.
- Microsoft Copilot: Included with Dartmouth's Microsoft 365 license. Usage limits are managed by Microsoft and are not shown to users as a visible token counter.
- ChatGPT, Gemini, and other consumer AI tools: Access levels and supported use cases vary by tool. Check the AI at Dartmouth site (ai.dartmouth.edu) for the current list of supported tools before using them with Dartmouth data.
Tips for Using Your Tokens Efficiently
- Be specific and concise in your prompts. Vague questions tend to produce longer, less focused answers that use more tokens.
- Ask for a specific output length, such as "in two sentences" or "as a five-item bulleted list," when a short answer will do.
- Trim documents before uploading. If you only need one section of a long file, share just that section instead of the whole document.
- Start a new conversation when you switch topics. Carrying over a long, unrelated conversation uses tokens without adding value.
What Happens When You Run Out of Tokens?
On Claude for Education, your token limit refreshes automatically every 5 hours, so you can simply wait and try again. On Dartmouth Chat, if you reach the daily credit limit, switch to a local model (Mistral or Llama) or wait until the limit resets the next day.
External Resources
See Related Articles to the right for more information. If you still have questions, please submit a ticket to ITC.