# token-optimization

Sabrina shares six practical ways to cut Claude Code token usage in half: switch models, clear context, use CLIs instead of MCP, compress MCP output, slim down CLAUDE.md, and route simple tasks to local models. They’re all free, and most take less than five minutes.

After analyzing Claude Code Pro usage reports, Lydia Hallie shared the four most effective ways to save tokens: use Sonnet by default, turn off extended thinking when you don't need it, don't resume idle sessions, and manually cap the context window.

Claude Code creator Boris Cherny cleared the air directly: every subscriber uses the same Opus 4.6 — there is no secret smarter model. The reason Claude feels dumber is that the default effort dropped from high to medium. One command brings it back.

Most token waste is invisible: Extended Thinking on tasks that don't need it, Opus handling work a Sonnet could do, context filling before you compact. ECC's token-optimization.md combines MAX_THINKING_TOKENS + model routing + strategic compact — author Affaan Mustafa says the savings reach 60-80%.

An AI Agent burns 34,500 tokens of system prompt every single conversation turn. The author used layered loading (always-on vs on-demand) plus a dual-model strategy to cut monthly costs from $568 down to $120-150 — a 75% reduction. Full breakdown with real numbers inside.