Claude Code: Models, Cost, and Context
Choosing the right model and effort level, understanding what fills the context window, and controlling what you spend
Claude Code gives you real levers over capability, latency, and cost: which model runs, how hard it thinks, whether it runs faster for more money, and how aggressively it manages its own context window. This subject teaches you those levers precisely — model aliases vs. pinned IDs, effort levels, fast mode, the advisor tool, prompt caching mechanics, subagent model routing, and the concrete habits that separate a $3/day session from a $30/day one — so you can make deliberate tradeoffs instead of accepting whatever the defaults hand you.
Practice questions (17)
-
View →
Choosing Model and Effort for Four Task Profiles
Intermediate · Free -
View →
What Actually Happens When You Edit CLAUDE.md Mid-Session
Intermediate · Free -
View →
Compaction Failure Modes on a Four-Hour Session
Advanced -
View →
Budgeting a 200-File Security Audit
Advanced -
View →
Designing an Organization Model Policy
Advanced