Feature Request
Current behavior:
Claude Code hard-caps the context window at 200K tokens for all users, including Max ($100/mo) and Max 20x ($200/mo) subscribers. This applies even when using Claude Opus 4.6, which supports up to 1M tokens via the API.
Requested behavior:
Increase the available context window for Max subscription users in Claude Code. Even an intermediate step to 400K tokens would be a meaningful improvement for users working on large codebases, long refactoring sessions, or complex multi-file tasks.
Why this matters:
- Max subscribers are paying a premium specifically for heavy usage of Claude's most capable models
- The underlying model (Opus 4.6) already supports 1M tokens — the 200K cap is a Claude Code limitation, not a model limitation
- The current 200K window (minus system prompt, tools, skills, and autocompact buffer) leaves only ~150K of usable space for actual conversation
- Larger codebases and longer sessions frequently hit this limit, triggering autocompact and losing valuable context
Related issues:
Suggested approach:
- A configurable context window size (e.g., via
claude config) would give power users control over the trade-off between context size and response quality/speed
- Even a modest increase to 400K–500K would significantly improve the experience for Max subscribers without requiring the full 1M beta
🤖 Generated with Claude Code
Feature Request
Current behavior:
Claude Code hard-caps the context window at 200K tokens for all users, including Max ($100/mo) and Max 20x ($200/mo) subscribers. This applies even when using Claude Opus 4.6, which supports up to 1M tokens via the API.
Requested behavior:
Increase the available context window for Max subscription users in Claude Code. Even an intermediate step to 400K tokens would be a meaningful improvement for users working on large codebases, long refactoring sessions, or complex multi-file tasks.
Why this matters:
Related issues:
Suggested approach:
claude config) would give power users control over the trade-off between context size and response quality/speed🤖 Generated with Claude Code