[BUG] HELLO prompt cost 15c (sonnet 4.6 ext) . Extra usage credits consumed abnormally fast — $20.69 (10%) drained within minutes on Max 20x plan -

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?

I'm on the Max 20x plan. When my plan usage limit was reached and it switched to Extra Usage credits, my credits were consumed at an abnormally high rate. Within approximately 5-10 minutes, ~$11.38 (6%) was consumed. The charges have since continued to climb rapidly — the extra usage is now at $20.69 spent (10% used) with a current balance of $179.30 out of a $200 monthly spend limit (resets May 1).

This rate of consumption does not match my actual usage. During normal plan usage, reaching 5-6% would typically take at least a day. But once Extra Usage kicked in, the same dollar amount was drained in minutes. I was using the same model throughout — nothing changed about my prompts or workflow.

Key observations:

• Each prompt in Claude Code (Mac Desktop) was costing $3-4 in extra usage credits
• - Pushing code to GitHub (no LLM involvement) caused extra usage to jump from $12.35 to $19.58
• - - The charges kept growing even with minimal interaction, now reaching $20.69
• - - - This issue only started after Extra Usage was enabled — plan usage consumption rate was normal

I contacted Anthropic support (Conversation ID: 215473777554326), but the AI agent was unable to investigate the specific charges, could not escalate to a human agent, and repeatedly gave the same generic response about token consumption at usage boundaries. No granular per-prompt usage breakdown is available on the Max plan to verify which requests consumed the credits.

Other users have reported similar issues on Reddit, including cases where a simple "hello" in a new session consumed 2% of usage.

What Should Happen?

Extra usage credits should be consumed at a rate consistent with the actual prompts being sent. If the plan usage meter takes a full day to consume 5-6%, the extra usage credits should not consume the equivalent dollar amount in minutes for the same type of prompts and model. Users should also have access to a granular per-request usage breakdown so they can verify what consumed their credits.

Error Messages/Logs

No error messages — this is a billing/usage tracking issue.
Usage dashboard currently shows:
- $20.69 spent (10% used)
- $179.30 remaining balance
- $200 monthly spend limit
- Resets May 1

Steps to Reproduce

1. Use Claude on a Max 20x plan until the plan usage limit is reached
2. 2. Enable Extra Usage credits
3. 3. Continue using Claude Code (Mac Desktop) with the same model and similar prompts as before
4. 4. Observe that extra usage credits are consumed at a drastically higher rate compared to plan usage — approximately $3-4 per prompt
5. 5. Even non-LLM actions (e.g., pushing code to GitHub while Claude Code is open) appear to trigger extra usage charges
6. 6. Extra usage went from $0 → ~$11.38 in ~5-10 min, and has now reached $20.69 (10%) with minimal interaction

Claude Model

Not sure / Multiple models

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

Claude 1.569.0 (49894a) 2026-04-02T20:01:42.000Z

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Other

Additional Information

• Anthropic support conversation ID: 215473777554326
• - There were resolved incidents with Sonnet 4.6 and Opus 4.6 around the same timeframe (March 31 - April 4) that may be related
• - - Similar issues reported by other users: see #43678, #43677, #41521, #38239, #41084
• - - - Auto-reload is off; monthly spend limit set to $200
• - - - - Current extra usage: $20.69 spent (10% used), $179.30 remaining balance, resets May 1