Anthropic · CCDV-F
Model selection, tokens, and cost
CCDV-F · Lesson 14 of 21 · 2:02:35
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the Claude Certified Developer – Foundations exam, on the exam guide version 1.0, effective July 2026. In this lesson: choosing among Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5 with your own evaluations, effort before model, breaking changes in migration guides, usage tracking and the token counting endpoint, cost models, max tokens and task budgets, and prompt caching breakpoints. Includes three original practice questions.
Preview the topics
- Driftwood moves its order agent from Opus 5 to Opus 5.5, and every request that forces one tool with tool_choice now fails with a 400 error. What should change?
- Driftwood’s research agent runs for many turns, and the team wants Claude to pace its work and finish gracefully within about 200,000 tokens for the whole task. Which feature fits?
- Driftwood marks a system prompt of about 300 tokens with cache_control on Sonnet 5.5, and the usage shows no cache writes and no cache reads, with no error. Why?