[BUG]

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?

Plan: Max
Platform: Windows
Issue: Cowork/Dispatch sessions hitting rate limits within minutes of starting — began after recent auto-update
Nothing changed on my end — same usage patterns that worked fine for 1-2 months
Suspected cause: Version silently switched to claude-opus-4-6[1m] for all Cowork sessions (see issue #33154) — tighter rate limits on 1M context are exhausting Max quota on trivial tasks
Request: Either revert the default model flag to standard context, or fix 1M context rate limits to be compatible with Max plan quotas

What Should Happen?

What Should Happen:
Cowork/Dispatch sessions on a Max plan should run reliably without hitting rate limits within minutes of starting a simple task. The expected behavior is consistent with how the product functioned for the previous 1–2 months of use — where the same usage patterns, same task complexity, and same workflow ran without interruption.
Specifically:

Cowork should default to the standard context model (claude-opus-4-6) rather than the 1M context variant (claude-opus-4-6[1m]) unless the user explicitly requires extended context
Sub-agents spawned by Cowork should not each independently request 1M context windows for routine tasks
A Max plan subscriber should be able to complete a simple Dispatch task without exhausting their 5-hour session quota in under 5 minutes
Rate limits should accurately reflect actual token consumption, not a ghost/corrupted state caused by a model flag regression introduced in a silent auto-update

The May 6 doubling of 5-hour limits has not resolved this — the root issue is the model being called, not the quota size.

Error Messages/Logs

Steps to Reproduce

<img width="1080" height="1872" alt="Image" src="https://github.com/user-attachments/assets/fc2da2bf-a179-4dcc-a285-5e0b934dc791" />

<img width="1080" height="1872" alt="Image" src="https://github.com/user-attachments/assets/0c710b12-f8e8-49d8-9b8a-47197bf83e19" />

<img width="1080" height="1872" alt="Image" src="https://github.com/user-attachments/assets/f2a33228-a843-4d97-9b25-0fe9c5284094" />

As you can see, I continued my task that needed to be done with Cowork, and I've already hit my limit. This is unheard of before the update. We need to fix this. Please immediately, help me!

Claude Model

Sonnet (default)

Is this a regression?

Yes, this worked in a previous version

Last Working Version

Here are the answers to all three fields: Last Working Version (Claude Desktop): Version 1.1.6679

Claude Code Version

N/A — using Cowork via Claude Desktop, not Claude Code CLI

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

_No response_