Study Claims Claude Users' Reasoning Compute Is Secretly Throttled

A 65-day analysis of 43,000+ Claude Code calls found 39% of calls received zero thinking tokens with a median of just 123, and 'The Inference Gap' research shows effort settings do not correspond to actual reasoning levels, indicating users are heavily throttled.

2026-09-20 ~ 2026-09-20 · 2 related posts

• 65-Day Analysis of 43K Claude Code Calls Finds Thinking Budgets Silently Slashed — IcyEase · 2026-09-20
• Research claims Claude users are heavily throttled, effort settings don't map to consistent reasoning levels — pbaylies · 2026-09-20