Excessive token consumption - 70% usage within one hour of normal use

Description

Token usage and limits are being consumed at an unexpectedly high rate. Within approximately one hour of normal usage (consistent with prior usage patterns), roughly 70% of the token budget is consumed.

Observed behavior

• Normal conversational usage (not unusually heavy or complex tasks) depletes ~70% of available tokens within ~1 hour
• This appears to be a recent change — the same usage patterns previously did not consume tokens at this rate
• No objective metrics available, but the pattern is noticeable and reproducible through regular use

Expected behavior

Token consumption for similar usage patterns should remain consistent over time, or any changes in consumption rates should be communicated to users.

Environment

• Date observed: 2026-03-30
• Model: Claude Opus 4.6 (1M context)
• Platform: macOS (Darwin 25.1.0)
• Client: Claude Code CLI

Additional context

This may be related to changes in how tokens are counted, context management, or model-level changes that increase token throughput per interaction. Would be helpful to have visibility into per-session token usage breakdowns.