[Feature Request] Enable 1M token context window for Claude Max subscribers

Feature Request

Allow Claude Max subscribers to utilize Sonnet or Opus full 1 million token context window capability in Claude Code.

Claude Code limits the context window to 200,000 tokens across all models (Opus 4.6, Sonnet 4.5, Haiku 4.5), regardless of the model's native capabilities. When approaching this limit, Claude Code's auto-compact feature compresses prior conversation history to stay within the window. This would be particularly valuable for Claude Max tier subscribers, who already have access to higher-tier models and could benefit from extended context in professional development workflows.

Justification

While auto-compaction is helpful for maintaining conversation continuity, it introduces two significant limitations:

1. Performance degradation: As the context fills, response times slow considerably due to the overhead of processing and compacting large conversation histories.
2. Loss of critical context: Auto-compaction necessarily discards or summarizes earlier conversation content, which can result in:

• Loss of nuanced technical details from earlier in the session
• Inability to reference specific code snippets, file contents, or decisions made earlier
• Reduced effectiveness for long-running debugging sessions or multi-phase implementation work
• Need to repeatedly re-explain context that was previously established

Expected Benefit

Leveraging the full 1M token window would:

• Eliminate premature context loss for extended development sessions
• Reduce performance degradation from constant auto-compaction
• Better support complex, multi-file refactoring and architectural work
• Align Claude Code's capabilities with what's available via the Anthropic API
• Provide additional value to Claude Max subscribers working on large codebases

Environment Info

• Platform: darwin
• Terminal: Apple_Terminal
• Version: 2.1.42
• Feedback ID: 0df5318c-19fb-491e-a629-d5eee1a813fb