[BUG] [Bug] 4x token consumption spike after silent Opus 4.6 rollout - no config changes

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Bug Description

After the Opus 4.6 release (Feb 5, 2026), my daily token consumption increased approximately 4x without any changes to my configuration, workflows, hooks, or CLAUDE.md files.

Evidence from /stats (Last 7 days, Models view):

| Period | Tokens/Day | Notes |
|--------|-----------|-------|
| Feb 4–6 | ~0 | Normal idle |
| Feb 6–7 | ~87k | First spike after Opus 4.6 appeared |
| Feb 8–10 | ~347k | 4x increase with same workload |

Model breakdown (last 7 days):

• Opus 4.6: 95.4% — In: 257.9k · Out: 288.8k
• Opus 4.5: 3.9% — In: 7.4k · Out: 14.7k
• Sonnet 4.5: 0.7% — In: 3.1k · Out: 1.1k

Key observations:

1. Output tokens (288.8k) exceed input tokens (257.9k), suggesting the model is generating significantly more verbose responses or consuming tokens via extended thinking
2. No changes were made to my setup — same projects, same hooks, same routines
3. The spike correlates exactly with the Opus 4.6 rollout date
4. This is consistent with issue #24243 (regression in per-turn token consumption)

Expected Behavior

Token consumption should remain consistent when user configuration has not changed. If the default model was silently upgraded to Opus 4.6, users should have been notified about the cost implications.

Questions

1. Was the default model automatically switched to Opus 4.6 without user opt-in?
2. Is extended thinking enabled by default on Opus 4.6, and does it count toward usage limits?
3. Are there known regressions in token efficiency between Opus 4.5 and 4.6?

Environment Info

• Claude Code version: 2.1.39
• Platform: macOS
• Plan: Pro Max 20x

<img width="550" height="457" alt="Image" src="https://github.com/user-attachments/assets/3da1e247-d92c-4da8-bebe-7fe293520ce6" />

• No changes to hooks, MCP servers, skills, or CLAUDE.md

What Should Happen?

Token consumption should remain consistent when user configuration has not changed. If the default model was silently upgraded to Opus 4.6, users should have been notified about the cost implications before the switch. Same workflow with same configuration should result in same token consumption levels.

Error Messages/Logs

Steps to Reproduce

1. Use Claude Code with a stable configuration (same hooks, CLAUDE.md, MCP servers, skills) for weeks/months
2. Claude Code auto-updates and silently switches default model to Opus 4.6 after Feb 5, 2026 release
3. Continue same workflow, same projects, same routines
4. Run /stats and observe 4x increase in daily token consumption (from ~87k to ~347k tokens/day)
5. Check model breakdown: Opus 4.6 accounts for 95.4% of usage with output tokens (288.8k) exceeding input tokens (257.9k)

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.39 (Claude Code)

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Cursor

Additional Information

_No response_