[BUG] 1M context window removed from Desktop Code tab model picker after v1.1.7714 update on Max plan

[EricPotratz]

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 updating Claude Desktop to v1.1.7714 (build 3bd6f6), the Opus 4.6 [1M] option no longer appears in the Code tab model dropdown on Max plan. Only "Opus 4.6" (200K), "Sonnet 4.6", and "Haiku 4.5" are shown.

The 1M option briefly flashes in the dropdown on launch, then reverts to plain "Opus 4.6". The /context command confirms a 200K ceiling.

Confirmed version regression: one machine had working 1M on the prior version with zero custom configuration. After updating to v1.1.7714, it disappeared immediately.

What Should Happen?

The model dropdown in the Code tab should show "Opus 4.6 [1M context]" for Max plan subscribers, as documented at code.claude.com/docs/en/model-config.

Error Messages/Logs

No error message. The 1M option silently disappears from the dropdown.
Also found stale s1mAccessCache in .claude.json with hasAccess:false from Dec 2025 that survived Reset App Data — this should be cleared by reset.

Steps to Reproduce

1. Have a Max plan account
2. Update Claude Desktop to v1.1.7714
3. Open Code tab
4. Click model dropdown — no 1M context option appears
5. On launch, "Opus 4.6 [1M]" briefly flashes then reverts to "Opus 4.6"

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

No response

Claude Code Version

1.1.7714 (3bd6f6)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Other

Additional Information

Tested on two Windows 11 machines, same Max plan account. Both affected on v1.1.7714.

Extensive local troubleshooting performed — none resolved it:

• Removed stale ANTHROPIC_API_KEY, ANTHROPIC_BASE_URL env vars
• Cleared s1mAccessCache from .claude.json (had hasAccess:false cached from Dec 2025)
• Removed oauthAccount field from .claude.json
• Replaced entire .claude.json with fresh auto-generated version
• Full Reset App Data + re-authentication
• Multiple Clear Cache and Restart
• Removed all ANTHROPIC env vars (verified at registry level)
• Cleaned settings.json of any model overrides

Note: s1mAccessCache persists through "Reset App Data" — this seems like a separate bug.

Likely related to #24208, #34143, #34333, #34435, #34593

[github-actions]

Found 3 possible duplicate issues:

1. Opus 4.6 shows 200K context instead of 1M on Claude Max plan (v2.1.75) #34143
2. [BUG] Opus 4.6 with 1M context is not available on Max 5x plan #34333
3. Max plan: 1M context window shows 200k, Remote Control unavailable #34083

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

[EricPotratz]

I think this issue is specific to Code tab within Claude Desktop. It was working until I updated this morning (March 19th)

[WUniq]

I'm furious with Anthropic. The exact same thing happened to me. I'm a MAX subscriber, I use Claude Desktop with code on Windows, and to top it all off, when I got the subscription I was in the middle of a major refactoring of the software I was working on. The update broke my session and ruined a lot of work I'd already done. I had to back up my code because when the tokens were reduced, Claude Opus 4.6 went haywire and corrupted the code. A big slap on the wrist for Anthropic.

[evanjcosgrove]

I've spent a hell of a lot of time debugging this and can add some specificity from my main.log that narrows the regression window:

2026-03-19 18:36:58 — LocalAgentModeSessionManager → model: claude-opus-4-6[1m]  ← working
2026-03-19 19:59:06 — LocalAgentModeSessionManager → model: claude-opus-4-6       ← [1m] gone
2026-03-20 16:04:28 — LocalAgentModeSessionManager → model: claude-opus-4-6       ← still gone

The [1m] suffix disappeared between 18:36 and 19:59 UTC on March 19 — which lines up exactly with the ["Elevated errors on Claude Opus 4.6" incident] that was marked resolved at 16:16 UTC that day.

Theory behind it: the incident resolution disabled or reset the 1M entitlement flag server-side for some subset of Max accounts.

What I've tried and hasn't worked (Claude desktop app + cowork on Mac)

• Clearing oauth:tokenCache from ~/Library/Application Support/Claude/config.json → no effect
• Setting ANTHROPIC_DEFAULT_OPUS_MODEL: "claude-opus-4-6[1m]" in ~/.claude/settings.json → only affects CLI, Cowork ignores it entirely (of course it does, but I'm trying everything at this point).
• Full logout + login → no re-evaluation of model entitlements (what a surprise).
• Quit app + relaunch → model stays at standard 200K (this was the first recommendation from Claude's useless support agent, so helpful).
• Reset App Data from the app menu → does not clear the stale entitlement cache (confirming @EricPotratz 's s1mAccessCache finding)

The supported-features-info.json in my logs still shows yukonSilver: supported and yukonSilverGems: supported, so the feature flags for Cowork itself are fine.. it's specifically the 1M model variant entitlement that's broken.

This needs a server-side fix. Either the s1mAccessCache needs to be reset for affected accounts, or the entitlement check that runs during LocalAgentModeSessionManager model resolution needs to be patched.

---

Hope this helps - ya'll are not crazy, this is very frustrating that we are paying $200/month for degraded service/features while Anthropic Support gaslights us like there's no issue.

[blumenkohlbinary]

Same issue here. Max plan, Claude for Windows v1.1.7714, Windows 11.

Timeline:

• Had 1M context working for ~2 days
• After the v1.1.7714 update on March 19, context dropped back to 200K
• tengu_hawthorn_window in .claude.json is set to 200000 by the server
• cachedExtraUsageDisabledReason: "org_level_disabled" keeps reappearing after deletion

What I've tried (none worked for Desktop Code tab):

• Deleting .claude.json (regenerated with 200K by server)
• Patching tengu_hawthorn_window to 1000000 (overwritten on next session)
• Removing cachedExtraUsageDisabledReason (comes back)
• Clearing Desktop caches (Cache, Session Storage, Local Storage, fcache)
• Full app restart
• Setting ANTHROPIC_DEFAULT_OPUS_MODEL=claude-opus-4-6[1m] in ~/.claude/settings.json — works for CLI but Desktop Code tab ignores it
• No DISABLE_TELEMETRY or CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC set anywhere

Confirming @evanjcosgrove's analysis — this is a server-side entitlement issue that needs to be fixed by Anthropic. Local workarounds cannot override the Desktop's LocalAgentModeSessionManager model resolution.

[leolln]

Same issue here

[stohn]

Even worse for me.

I can not continue all the different COWORK sessions (projects) I used when the 1M context was available.
Whatever I type in there brings the "Prompt to long" error and I have no options.

Now I wait and hope the 1M context window for Cowork (with Max 20x Plan) comes back quick !!!

And boy did the 1M context made a difference for me. Now with the much smaller window claude feels like he forgets what he did 20 minutes ago frequently. This basically never happend with the 1M window.

I WANT IT BACK... PLEASE :-)

[tomsouza4]

Same issue on Mac!