RealFeedApp Opus 5.5 saves tokens with fewer turns
Although the price per token is identical, Opus 5.5 completes tasks in fewer turns, reducing total cost. Fewer iterations mean less context resending and lower overall expenditure.
- Field
- applying language models
- What they did
- Compared the cost of executing tasks on Opus 5.5 versus Opus 5, demonstrating that token consumption depends on task complexity rather than just the price per token.
- Why it matters
- Users need to understand that attempts to reduce token usage (such as by reducing context) can lead to the need for retries, ultimately making the task more expensive.
#opus#token cost#llm#api pricing#inference
