Table of Contents
Introduction (Required)
KeanU AI provides access to multiple AI models, including OpenAI (GPT) and Anthropic (Claude). Each user has a monthly token allocation used when interacting with AI models.
Understanding how tokens are consumed can help users maximize their allocation while still obtaining high-quality results.
Use this solution if you:
- Want to understand how AI tokens are consumed.
- Need to check your current token usage.
- Want to maximize your monthly token allocation.
- Need guidance on selecting the appropriate AI model for your task.
- Have questions about token resets or usage limits.
Prerequisites
Before using this solution, ensure you:
- Have access to KeanU AI via myapps.kean.edu. All Kean students, faculty and staff are able to access KeanU AI.
- Are signed in with your Kean credentials.
What Are Tokens?
Tokens are the units used to measure AI usage.
Examples of token consumption include:
- Questions and prompts you submit
- AI-generated responses
- Uploaded documents that are analyzed
- Extended conversations and chat history
Your token balance is deducted based on the amount of processing required for each request.
Note: Token allocations reset on the 1st of each month.
Choosing the Right AI Model
Selecting the appropriate model can help conserve tokens while still achieving your desired results.
Check Your Token Usage
To view your remaining token allocation:
- Sign in to KeanU AI.
- Locate the usage indicator in the upper-right corner.
- View the token usage circle to view your current consumption.
The usage indicator displays the percentage of your monthly allocation that has been used. The circle color will change to red when your allocation token is almost finished for the month.
Switching Between AI Models
KeanU AI allows users to switch between available AI models based on the type of task they are working on. Users can switch between available AI models only when using OneChat.
To Change Models in OneChat:
- In your KeanU AI Chat, locate the Model Selector at the top of the chat window.
- Select the current model name to view available options.
- Choose the model that best fits your task.
- Continue your conversation using the selected model.
Personal Agents do not support model switching during a conversation. Personal Agents use GPT-based models, and the model cannot be changed while chatting with the agent.
If you want to use a different GPT model for a Personal Agent, you must:
- Open the Personal Agent for editing.
- Select the Edit icon at the upper right corner.
- Under the Model dropdown, select Change, choose your desired model, and then select Save to apply the update.
- The select the "Update" button to update your Personal agent.
The newly selected model will be used for future conversations, but users cannot switch models directly from the chat interface as they can in OneChat.
Reduce Token Consumption
To make the most of your monthly allocation:
Be Specific
- Instead of broad requests, provide clear instructions.
- Less Efficient: "Tell me everything about artificial intelligence."
- More Efficient: "Provide a 5-bullet summary of how AI can improve higher education."
Upload Only Necessary Content
Large documents consume more tokens.Consider:
- Uploading relevant sections only.
- Removing unnecessary pages.
- Breaking large files into smaller documents.
Use GPT First
- For most routine tasks, GPT provides effective results while conserving tokens.
- Use Claude when additional reasoning or document analysis is required.
Avoid Repeating Information
- Continue working within an existing conversation when possible instead of re-entering the same context multiple times.
How to switch between models
Frequently Asked Questions
- Why am I out of tokens?
- You may have:
- Uploaded large documents
- Used advanced reasoning models extensively
- Conducted lengthy conversations
- Generated large amounts of content
- You may have:
- When do my tokens reset?
- Token allocations automatically reset each month.
Related Articles
Contact Information
- KeanU AI Support Email: [email protected]
- AI Misues email: [email protected]