Customer-Hosted Deployment Codebeamer AI Token Usage
A token represents a unit of text processed by the large language model (LLM). Tokens are consumed for:
• Input prompts sent to the model.
• Cached input reused during an AI task.
• Output generated by the model.
The total number of tokens consumed depends on factors such as:
• The AI assistant being used.
• The number of LLM calls required to complete a task.
• The size of input and output text per call.
 |
Token consumption applies only to customer‑hosted Codebeamer AI. PTC‑hosted Codebeamer AI uses AI credits instead of tokens.
|
Token consumption by AI Action
The following list shows an example of average token consumption per AI action for common Codebeamer AI assistants in customer-hosted deployment. These values are indicative and may vary depending on context size and usage patterns.
• Requirement Assistant—
◦ Approximately 40,000 tokens per action to evaluate a requirement.
◦ Approximately 22,000 tokens per action to rewrite a requirement.
• Test Case Assistant—Approximately 13,140 tokens per task.
• Search Assistant—Semantic search uses only a small number of tokens per query. For example, approximately 75 tokens per search.
• AI Indexing—Approximately 1.5 million token per million indexed words.