Disabling Claude's 1M Context Slashes Token Usage, Tests Show
On September 2, developer dotey posted a series of tips for reducing Claude Code token consumption, centered on disabling the 1M context window.
Confirmed
- Set the environment variable CLAUDECODEDISABLE1MCONTEXT to "1" in /.claude/settings to disable the 1M context and let some models run with a 200K window budget
- dotey reported that the setup can save a Claude account roughly 15 million tokens per month
- A user compared two accounts on two machines: the one with 1M context enabled burned tokens visibly faster, and consumption dropped immediately after disabling it
- The /autocompact auto or /autocompact 200k command adjusts the auto-compaction threshold; it can also be set to 500K to keep context longer before compacting. The tester found /autocompact 200k the simplest option
- dotey noted the effect is especially noticeable when using the Fable 5.1 model
Why it matters
- The 1M context raises token consumption and cost significantly; for most everyday tasks, a 200K window with auto-compaction suffices
- These settings take effect with a single variable or command, offering a low-effort way to control Claude subscription quotas and API costs
2026-09-02 ~ 2026-09-02 · 5 related posts
Primary sources
- [source] Tip: Disabling Claude 1M context saves tokens — dotey · 2026-09-02
- Claude Code tip: tune context compaction with /autocompact — dotey · 2026-09-02
- [source] Test: Claude 1M Context Window Significantly Increases Token Usage — dotey · 2026-09-02
- [source] Slash Claude costs: How to save 15M tokens monthly with /doctor and config tweaks — dotey · 2026-09-02
1 near-duplicate retellings: dotey