Body
Overview
This article describes how CCBC uses tokens within Microsoft Copilot based on our Enterprise contract.
Audience
Faculty, Staff and Students
Description
Based on the Maryland Enterprise Education Consortium (MEEC) contract, CCBC has access to Microsoft's version of Chat-GPT, Microsoft Copilot. Copilot's usage of tokens is consumed in a different manner than the way Chat-GPT's token usage is for the public models.
Version Available to Full-time faculty and staff and All students
Microsoft 365 Copilot (General Chat) :
- General Chat is available to all full-time faculty and staff and all students.
- Tokens are relevant for technical limits, not direct billing.
- Copilot counts both input and output tokens for billing purposes. For example, a long prompt and a detailed response consume more tokens.
- Tokens are units of text processed by the underlying language model. Roughly 1,000 tokens ≈ 750 words.
- Token limits vary by model:
- GPT‑4o: up to 128,000 tokens input and 16,384 tokens output per request.
- GPT‑5 migration is ongoing, but similar token-based accounting applies.
Versions that require Additional License/Cost and not generally available to CCBC faculty, staff and students
Microsoft 365 Copilot (Advanced Chat):
- Advanced Chat Features includes Microsoft developed Retrieval Augmented Generation (RAG) Agents
- Researcher - Works the same as general chat but documents and annotates and cites all sources being used.
- Analyst - Takes larger forms of data and performs analytics and produced outcomes.
- Pricing is seat-based (Currently $30/user/month for Enterprise plans), not per token. Tokens are relevant for technical limits, not direct billing.
- Advanced Chat also includes the ability to create custom agents.
- Advanced Chat includes all of the features and follows the same token usage as General Chat.
Copilot Studio & Agent Scenarios:
- Uses Copilot Credits as the common currency. Credits correlate with token usage and are available via:
- Prepaid packs (annual or monthly).
- Pay-as-you-go meters billed through Azure.
- Pre-purchase plans for discounts.
Microsoft Fabric Copilot (Power BI, Data Factory, etc.):
- Charges are based on Capacity Units (CU) per 1,000 tokens:
- Input: 100 CU-sec per 1,000 tokens.
- Output: 400 CU-sec per 1,000 tokens.
- Example: A request with 2,000 input tokens + 500 output tokens = ~6.67 CU-minutes.
GitHub Copilot Enterprise:
- Group plans include premium request quotas (e.g., 1,000 requests/user/month). Additional requests cost extra, but pricing is per seat, not per token.