Context Window Calculator

1. Enter the model window
Use the token limit for the exact model and endpoint.

2. Add system and tool tokens
Include instructions, schemas, and framework-generated messages.

3. Measure conversation history
Count the messages that will be sent with the next request.

4. Add retrieved context
Include document chunks, citations, metadata, and separators.

5. Reserve output tokens
Leave enough capacity for the expected completion.

6. Set a safety margin
Protect against estimation error and variable tokenization.

7. Check remaining capacity
A negative result means the planned request is too large.

Remaining context = context window − system tokens − history tokens − retrieval tokens − reserved output − safety margin tokens

Where:

• Safety margin tokens: context window × safety margin percentage
• Planned token use: system + history + retrieval + reserved output
• Window utilization: (planned use + safety margin) ÷ context window × 100

Assumptions: All components share one context budget. The calculator assumes the stated context window applies to combined input and output.

What the result means

Use the primary result together with the detailed breakdown. Scenario assumptions can materially change the estimate.

This planning tool does not replace provider documentation, a production benchmark, or professional advice.

Given: A 128,000-token window, 3,000 system/tool tokens, 20,000 history tokens, 50,000 retrieved tokens, 4,000 reserved output tokens, and a 10% safety margin.

Calculation:
Planned use = 3,000 + 20,000 + 50,000 + 4,000 = 77,000 tokens.
Safety margin = 128,000 × 0.10 = 12,800 tokens.
Remaining context = 128,000 − 77,000 − 12,800 = 38,200 tokens.

Result: About 38,200 tokens remain available before the safety buffer is consumed.