[BUG] Max plan: context window stuck at 200K instead of 1M (hasExtraUsageEnabled not updated after Pro → Max upgrade)

[phdmrruslan-droid]

Preflight Checklist

• I have searched existing issues and this hasn't been reported yet
• This is a single bug report (please file separate reports for different bugs)
• I am using the latest version of Claude Code

What's Wrong?

After upgrading from Pro to Max 5x plan on March 21, 2026, Claude Code context window remains at 200K instead of the expected 1M. The /context command shows 65k / 200k tokens (32%). The organization-level flag hasExtraUsageEnabled in .claude.json is stuck at false with cachedExtraUsageDisabledReason: "org_level_disabled" — inherited from the original Pro plan authorization on March 8, 2026, and never updated after the Max upgrade. Manually editing the file has no effect — the server overwrites it back to false on every launch.

What Should Happen?

Context window should show Xk / 1000k for Max 5x plan subscribers. The hasExtraUsageEnabled flag should be true at the organization level.

Error Messages/Logs

From .claude.json (oauthAccount block):
"hasExtraUsageEnabled": false

From .claude.json (top level):
"cachedExtraUsageDisabledReason": "org_level_disabled"

/context output:
Tokens: 65k / 200k (32%)

Steps to Reproduce

1. Create account on Pro plan (Feb 25, 2026)
2. Authorize Claude Code for the first time while on Pro (Mar 8, 2026) — organization is created with extraUsage: disabled
3. Upgrade to Max 5x plan (Mar 21, 2026)
4. Run /context in Claude Code — shows 200k limit instead of 1000k
5. Run /logout then /login — reauthorize — limit still 200k
6. Manually edit .claude.json to set hasExtraUsageEnabled: true — server overwrites it back to false on next launch

Claude Model

Opus

Is this a regression?

No, this never worked

Last Working Version

No response

Claude Code Version

2.1.101

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

Organization UUID: aa8ba640-d6eb-46d5-86d8-4e1900f9ea3b
Account UUID: cdc0e8e9-84cb-4c79-b731-494a2bf99b19
Plan: max_5x_monthly (Stripe subscription, active until 2026-04-21)
Email: ph.d.mr.ruslan@gmail.com

Timeline:

• 2026-02-25: Account created (Pro)
• 2026-03-08: First Claude Code auth (org created with extraUsage: disabled)
• 2026-03-21: Upgrade to Max 5x (plan updated, org flag NOT updated)
• 2026-04-12: Bug discovered

Support ticket via Fin AI Agent (Conversation ID: 215473870469578) — bot looped without escalation.

Request: Please update hasExtraUsageEnabled to true for the organization listed above.

[github-actions]

Found 2 possible duplicate issues:

1. [BUG] CLI caches subscriptionType and rateLimitTier indefinitely — plan upgrades not reflected #43639
2. [Bug] Regression: Claude Opus context window downgraded from 1M to 200K on Max plan #44403

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

[phdmrruslan-droid]

This is related but NOT a duplicate. Issues #43639 and #44403 may describe similar symptoms, but my case has specific diagnostic details: the root cause is a stale org-level hasExtraUsageEnabled: false flag that was never updated after a Pro → Max upgrade on March 21, 2026. I have confirmed that the server overwrites local .claude.json edits, proving this is a server-side issue requiring an org-level flag update for organization UUID aa8ba640-d6eb-46d5-86d8-4e1900f9ea3b.

See also related issues: #43872, #39841 — this appears to be a systemic problem affecting Max plan subscribers.

[nathanhouse]

Confirming this is still happening as of today (April 13, 2026) and appears to be a broader server-side regression, not just a Pro → Max upgrade caching issue.

Setup:

• 3 separate Max 20x ($200/mo) accounts — all affected
• Claude Code v2.1.104 (also reproduced on v2.1.87)
• macOS, native install

Diagnostics:

• tengu_hawthorn_window in cachedGrowthBookFeatures: 200000 (should be 1000000)
• hasExtraUsageEnabled: false
• cachedExtraUsageDisabledReason: "org_level_disabled"

What I've tried (none fixed it):

• Updating from v2.1.87 → v2.1.104
• Logging out and back in (/login) to force entitlement refresh
• Setting CLAUDE_CODE_AUTO_COMPACT_WINDOW=1000000 env var (doesn't override the API-level context limit)
• Switching between all 3 Max accounts — all show the same 200K cap

This was working fine on 1M until ~30 minutes before I noticed (~19:30 CEST today). All 3 accounts dropped simultaneously, which points to a server-side flag change rather than per-account caching.

[s1lentblade]

Swap to model:
/model claude-opus-4-6[1m]

Persistent default model:
Add "model": "claude-opus-4-6[1m]" to settings.json
Or setup env variable ANTHROPIC_MODEL=claude-opus-4-6[1m]

/context shows 1M and context window certainly extends past 200K,
Current plan Claude Max 5x

Summary:
Probably was a flag change to claude-opus-4-6[1m] to use the 1M context model where it otherwise defaults to the 200K model if you don't know the exact flag (perhaps to cut down on quantity of users using the 1M model?)

[Kt0k0lwiek]

Same issue here. Max plan, Claude Code with claude-opus-4-6. The context_window_size field in the statusline JSON payload reports 200000 instead of 1000000. Settings have autoCompactWindow: 1000000 configured correctly. The context was showing 1M yesterday, reverted to 200K today without any configuration change on my end.