[BUG] Context limit reached at ~200K tokens despite 1M context model

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?

Running Claude Code v2.1.75 with Opus 4.6 (1M context) on Claude Max plan. The session hit "Context limit reached · /compact or /clear to continue" at approximately 197K tokens used (20% of advertised 1M context).

Running /context immediately after the limit message confirmed:
claude-opus-4-6[1m] · 197k/1000k tokens (20%)

The model correctly shows [1m] suffix and the denominator is 1,000k, but the practical limit appears to still be enforced around the old 200K boundary.

Environment:

• Claude Code v2.1.75
• Model: claude-opus-4-6[1m] (Opus 4.6 with 1M context)
• Plan: Claude Max
• Platform: Linux (WSL2) 6.6.87.2-microsoft-standard-WSL2
• Session type: Long design review conversation with multiple file reads and edits

What Should Happen?

Session should continue until closer to 1M token limit.

Actual behavior: Session stopped at ~197K tokens with "Context limit reached" message, despite 80% of context remaining unused.

Error Messages/Logs

Context limit reached · /compact or /clear to continue                 
                                                                                                                                                 
  /context output at time of error:                                                                                                              
  ⛁ ⛁ ⛁ ⛀ ⛁ ⛁ ⛀ ⛁ ⛁ ⛁ ⛁ ⛁ ⛁ ⛁ ⛁ ⛁ ⛁ ⛁ ⛁ ⛁   claude-opus-4-6[1m] · 197k/1000k tokens (20%)                                                        
                                                                                                                                                 
  Estimated usage by category:                                                                                                                   
  ⛁ System prompt: 5.9k tokens (0.6%)                                                                                                            
  ⛁ System tools: 9.1k tokens (0.9%)                                                                                                             
  ⛁ Custom agents: 956 tokens (0.1%)                                                                                                             
  ⛁ Memory files: 11k tokens (1.1%)                                                                                                              
  ⛁ Skills: 1.2k tokens (0.1%)                                                                                                                   
  ⛁ Messages: 170.3k tokens (17.0%)                                                                                                              
  ⛁ Compact buffer: 3k tokens (0.3%)                                                                                                             
  ⛶ Free space: 799k (79.9%)                                                                                                                     
                                                                                                                                                 
The limit appears to be enforced at the old ~200K boundary despite the model being configured for 1M context. The /context command itself correctly reports the 1M limit (1000k denominator) but the session termination threshold was not updated to match.

Steps to Reproduce

1. Using Claude Code v2.1.75 with Claude Max plan (Opus 4.6, 1M context)
2. Start a long conversation involving multiple file reads and edits (design document review session)
3. Model correctly identifies as claude-opus-4-6[1m] throughout the session
4. After approximately 197K tokens of conversation, receive "Context limit reached · /compact or /clear to continue"
5. Run /context immediately after the limit message
6. Output confirms only 20% of context used: "claude-opus-4-6[1m] · 197k/1000k tokens (20%)"
7. Session cannot continue despite 800K+ tokens remaining

Claude Model

Opus

Is this a regression?

No, this never worked

Last Working Version

_No response_

Claude Code Version

2.1.75 (Claude Code)

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

VS Code integrated terminal

Additional Information

_No response_