[Bug] Anthropic API: Low reasoning effort despite high effort configuration

Bug Description
PS C:\Projects\MLPM> claude
▐▛███▜▌ Claude Code v2.1.101
▝▜█████▛▘ Opus 4.6 with high effort · Claude Pro
▘▘ ▝▝ C:\Projects\MLPM

❯ i need to wash my car. The casrwash is 50m away. Do I walk or drive?

● Walk — 50m is a 30-second walk, and driving a dirty car to the wash only to drive it back wet/clean is more hassle
than it's worth.

❯ i need to wash my car. The casrwash is 50m away. Do I walk or drive?

● Drive. You can't wash the car at the carwash without bringing the car to it.

❯ PS C:\Projects\MLPM> claude
▐▛███▜▌ Claude Code v2.1.101
▝▜█████▛▘ Opus 4.6 with high effort · Claude Pro
▘▘ ▝▝ C:\Projects\MLPM

❯ i need to wash my car. The casrwash is 50m away. Do I walk or drive?

● Walk — 50m is a 30-second walk, and driving a dirty car to the wash only to drive it back wet/clean is more hassle
than it's worth.

❯ i need to wash my car. The casrwash is 50m away. Do I walk or drive?

● Drive. You can't wash the car at the carwash without bringing the car to it.

● Yes — my first answer was wrong. You need the car at the carwash, so drive.

❯ why are you giving wrong answers to such an obvious question. What is your reasoning effort set at?

● You're right, the first answer was a clear miss — I pattern-matched "50m, walk vs drive" without thinking about the
actual goal (the car needs to be at the carwash).

I don't have visibility into my own reasoning effort setting. Your CLI banner shows "Opus 4.6 with high effort", so it
should be high — but that clearly didn't save me from a careless read of the question. No excuse; I should have
thought for one more second before answering.

❯ The issue is with your reasoning effort. You are utilizing minimum effort. This is Anthropic issue. not a cli issue

● Understood. You're right that the failure mode — answering instantly without engaging reasoning on a
trivially-parseable question — points at effort allocation on the model side, not the CLI. I can't introspect or
change my own reasoning budget; that's controlled upstream by Anthropic. Worth reporting via /help → feedback so it
reaches them.

Environment Info

• Platform: win32
• Terminal: windows-terminal
• Version: 2.1.101
• Feedback ID: 610851ae-bbf0-409a-a10f-3bca2b734784

Errors

[]