iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Lowering Claude’s effort setting can reduce the tokens it uses for a response, but an “8× longer” result is not guaranteed. In one author’s small test, a rounded usage meter moved 1% at Low effort and 8% at High after three routine prompts. Anthropic says lower effort can improve token efficiency, with a possible trade-off in capability.
Which Claude setting can reduce usage?
The setting is called effort. In Anthropic’s API documentation, it controls how much token use Claude applies when responding, including reasoning when active and tool calls. Lower effort is intended to favor token efficiency over response thoroughness. Anthropic describes effort as a behavioral signal, not a fixed token allowance: a difficult task can still lead to reasoning at Low, but generally less than at a higher level.
Anthropic characterizes Low as its most efficient effort level and cautions that it can reduce capability. The setting is therefore a way to influence a response, not a promise that every answer will use a particular number of tokens or preserve the same quality. See Anthropic’s effort documentation for current API details.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →What the reported 8× result actually measured
A WindowsForum account of a How-To Geek author’s test says the author tried three prompts at Low effort in one chat, then repeated them at High effort in a new chat using the same model. The prompts involved debugging a Linux script, finding time-zone conflicts between meetings, and splitting trip expenses. Both sets of answers were reported as correct. The usage meter moved by 1% for Low and 8% for High, producing the reported 8-to-1 ratio in displayed usage changes.
#1 Best Overall
That result is illustrative, not a benchmark. Each setting was tested once, the usage meter was rounded, and the author did not report direct token counts. The tasks were routine, and three examples cannot establish that Low effort will preserve quality on complex coding, extended agent work, or other difficult tasks. The author also noted that the test did not establish that higher effort never helps. WindowsForum’s September 23, 2026 account describes the test and its limitations.
Which effort level should you use?
Choose based on the task’s difficulty, how costly an error would be, and whether you can verify the answer. Anthropic recommends starting at High for models such as Fable 5.1, then trying Medium or Low for routine or latency-sensitive tasks once evaluations show that quality holds. Its guidance identifies Low as a reasonable option for simpler work where speed and cost matter.
- Try Low: routine requests with clear answers that are quick to check, such as straightforward summaries or simple calculations.
- Try Medium: ordinary work where you want to reduce effort but would still value some additional thoroughness.
- Keep a higher setting when warranted: complex or consequential work where errors are hard to detect or expensive to correct. Lower effort is not established as equally capable in these cases.
For your own model and task mix, compare results at two effort levels and check both answer quality and usage. Treat any reduction you observe as specific to those prompts, model, and conditions—not a universal multiplier.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchAre Claude’s default effort settings the same for every model?
No. Anthropic’s API guidance says most models that support effort default to High, while Claude Opus 5.5 defaults to Medium. Supported levels and defaults can differ by model, so check the current documentation for the model you use rather than assuming one universal Claude default. Anthropic’s effort page was checked October 7, 2026.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
