claude-code - 💡(How to fix) Fix Opus 4.7 ([1m] variant) capped at 200K context in Claude Code Desktop on Max plan [1 comments, 2 participants]

---

Summary

On the Max plan, selecting Opus 4.7 with the explicit 1M variant ID (claude-opus-4-7[1m]) in Claude Code Desktop on Windows shows a 200K context window in /context instead of the expected ~1M. Previously had 1M on this same account/build before a temporary downgrade to Pro and re-upgrade back to Max.

Environment

Claude Code (CLI version on this machine, from claude --version): 2.1.80
Surface: Claude Code Desktop app on Windows 10
OS: Windows 10
Plan: Max (re-upgraded from Pro recently — Downgrade happened on approx April 4th. Upgrade happened May 1st/2nd (was around midnight))
Observed

Fresh session after a full Claude Desktop restart.
/model claude-opus-4-7[1m] accepted: Set model to claude-opus-4-7[1m].
/context reports: Model: claude-opus-4-7[1m], Tokens 153.3k / 200k (77%), Autocompact buffer 33k. Total cap is 200K, identical to the non-[1m] model.
Switching back to plain claude-opus-4-7 shows the same 200K cap.
/bug and /feedback are not registered slash commands in Claude Desktop, so this issue is filed via GitHub instead of in-app.
/model in Desktop accepts a model ID inline only; it does not surface an "Opus (1M context)" picker option that the CLI documentation describes.
~/.claude/settings.json had "model": "claude-opus-4-7" (plain ID); changing this to "claude-opus-4-7[1m]" does not change the /context cap either.
Expected

Per https://code.claude.com/docs/en/model-config.md#extended-context, on the Max plan Opus auto-upgrades to a 1M context window. /context should reflect ~1M when claude-opus-4-7[1m] is the active model.

Reproduction

On Max plan, open Claude Desktop on Windows.
/model claude-opus-4-7[1m] → accepted.
/context → shows 200K cap.
Quit Claude Desktop fully, relaunch, start a fresh session.
Repeat steps 2–3 → still 200K cap.
History (pre-bug, possibly relevant)

Account had 1M Opus context working previously on this same machine/build. Plan was downgraded to Pro temporarily, then re-upgraded back to Max. The 200K cap persists since the re-upgrade.

Already tried

Restart of Claude Desktop, fresh session.
Explicit claude-opus-4-7[1m] selection via /model.
Editing ~/.claude/settings.json to pin the [1m] variant.
Hypothesis

The 1M context entitlement is gated on cached account/plan state that wasn't refreshed after the Pro→Max re-upgrade. The model ID is recognized (selection succeeds), but the per-session context cap is still being computed from the prior plan tier.