[Bug] 1M context incorrectly requires extra usage on Max plan

I'm on Max 5x. I'm using the Claude co-worker (web, claude.ai/code). The model selector at the bottom shows "Sonnet 4.6" - standard, not 1M.
Every single message returns:
API Error: Extra usage is required for 1M context · enable extra usage at claude.ai/settings/usage, or use --model to switch to standard context
The error tells me to "use --model to switch to standard context" but there is no way to do that in the co-worker web UI. The model dropdown already shows Sonnet 4.6, not a 1M va
riant. There is no CLI flag available in the web interface.

I'm paying $100/month for Max 5x. I shouldn't need to pay additional money on top of that to use a core feature of the product.

Same issue for me, using Max 5x. Had subagents spawned while mowing the yard, came back to over +/-$25 in overages for a function that is explicitly covered in my plan already.

This is happening to me on Max 20x.

Same issue here. Workaround: buildproven/claude-kit#15 Deleting the model field from the front matter in SKILL.md fixes it.

Worked

same issue

Reproducing on Max (20x) — Opus 4.7 [1m], blocking all 1M usage

Confirming this bug is still live as of 2026-05-19 on claude-opus-4-7[1m].

Environment

• Plan: Claude Max (20x). Plan usage low (current session ~19%, weekly ~11% — well within limits).
• Model: claude-opus-4-7[1m] in Claude Code.

Symptom

• Every turn fails with: API Error: Extra usage is required for 1M context · enable extra usage at claude.ai/settings/usage, or use --model to switch to standard context.
• The 1M window for Opus is supposed to be included in Max with no extra usage (per docs and prior issues #46780, #47054, #45847). Instead it is being charged against pay-as-you-go usage credits.
• Consequence: usage credits hit their cap and the 1M context is now fully blocked for a Max subscriber who should have it included. Switching the session out of [1m] is the only workaround, which defeats the purpose of the Max 1M entitlement.
• Also reproduces via skills with model: frontmatter invoked from an Opus [1m] session (resolves to sonnet-4-6[1m]), matching #45847.

Expected: Opus 1M context on Max should not consume extra-usage credits.

#46780 and #47054 were closed but the behavior persists. Please re-investigate — this is making the advertised Max 1M entitlement unusable.

I received this email response from support.

Thanks for the report. This is a confirmed issue and the team is working on it. There isn't a workaround on your end at the moment, unfortunately; the fix has to come from our side, although I don't have a specific timeline I can promise.
We'll close this out for now but let us know if you're still seeing issues and we'll trace it further.

Same issue here on Max plan when invoking the academic-research-skills:ars-revision skill from Imbad0202/academic-research-skills (v3.9.4.2).

Environment:

• Claude Code: 2.1.146
• Plan: Claude Max
• Default model: Opus 4.7 with 1M context (Max default)
• OS: WSL2 / Ubuntu

Reproduction:

/academic-research-skills:ars-revision @some_draft.md
  ⎿ Read some_draft.md (492 lines)
  ⎿ API Error: Usage credits required for 1M context · run /usage-credits to turn them on, or /model to switch to standard context

What I verified before reporting:

• The command file originally had model: sonnet in its frontmatter. I removed it (per the workaround pattern in buildproven/claude-kit#15 and angie/agentic-setup@71cdcb3).
• After the fix, grep -rn "^model:" ... confirms no remaining sonnet pins anywhere in the plugin tree (other commands use model: opus or model: inherit).
• Parent session is confirmed on Opus 4.7 + 1M via /status.
• Claude Code fully restarted after the edit.
• Error persists.

So this is hitting the upstream bug cleanly — no plugin-side workaround can resolve it once the skill correctly inherits Opus 4.7 [1m] from the parent. The billing gate trips even on a model+plan combination that should be plan-covered per the Help Center documentation.

I have the same issue on Pro plan. I got this on evening, but after all night and recharging limits I also see this message after every prompt.

Behavior:
Every turn fails with: API Error: Extra usage is required for 1M context · enable extra usage at claude.ai/settings/usage, or use --model to switch to standard context.

Environment:

• Version: 2.1.147
• Plan: pro
• model: sonnet
• OS: windows

Claude Code fully restarted after the edit. But nothing happened.

I have the same issue. Changing the model for a basic model, fix the error 😢

same issue getting

Same issue with the Max plan. Try logging out and logging back in again, it should work.

same issue here using Max (20x):

Error: Error during compaction: API Error: Usage credits required for 1M context · turn on usage credits at claude.ai/settings/usage, or use --model to switch to standard context.

@danielramireze have you tried my solution? Log out and login again (/logout and /login command)

I'm seeing the same issue, as had been working in Opus 4.7 1M for a while, and am seeing the error when trying to switch back to Sonnet 4.6 on the Max 5x plan, even though all of my usages are very low currently.

Trying to run /compact when on Sonnet 4.6 shows an error related to usage as well.

Error: Error during compaction: API Error: Usage credits required for 1M context · turn on usage credits at claude.ai/settings/usage, or use --model to switch to standard context

Can still use both Opus model entries normally. The previously mentioned logout/login idea did not help.

> Same issue here. Workaround: buildproven/claude-kit#15 Deleting the model field from the front matter in SKILL.md fixes it.

This fixed my issue! Thank you!

@danielramireze have you tried my solution? Log out and login again (/logout and /login command)

Yes it worked

I'm hitting the same issue. This really is unacceptable. This is the advice I got from Claude.

this is a known bug, not a legitimate billing issue. Here's the full picture:
What's happening: Anthropic's official communication explicitly states that 1M context is included in Max, Team, and Enterprise plans with no extra usage required — but Cowork is forcing the claude-opus-4-6[1m] model flag, and the 1M context rate limits aren't properly compatible with Max plan quotas, meaning even simple tasks get blocked. GitHubGitHub
This has been a recurring Cowork-specific bug. The 1M context window was working correctly in Cowork as recently as late March 2026, then broke — and notably Claude Code on the same account still has full 1M context. It's a Cowork-only regression. GitHub
There's also a specific variant that may match your situation exactly: when a skill specifies model: sonnet in its frontmatter from an Opus 4.6 [1m] session, Claude Code resolves it to claude-sonnet-4-6[1m], inheriting the 1M context window. Sonnet 4.6 [1m] may require extra usage on Max, whereas Opus 4.6 [1m] does not. Your SAM.gov project uses skills (including your interview-guide and candidate-submission skills) which may be triggering exactly this. GitHub

<img width="846" height="797" alt="Image" src="https://github.com/user-attachments/assets/657197f3-18ae-496d-8f8f-ed35546fbc21" />

Same issue, was on Pro and upgraded then started hitting these errors sporatically, no warning it was over hte limit and was consuming extra credits, then would start erroring again.

Pretty disappointed I upgraded and then hit an issue that's making getting things done harder.

<img width="376" height="170" alt="Image" src="https://github.com/user-attachments/assets/352d8d16-9ecf-4cea-9f71-f561fef073ef" />

Experiencing the same issue on Max plan using Cowork. Session allotment was only at 27% when the error appeared. The 1M context error is blocking my work entirely. Please prioritize a fix for Max plan users.

the same here I'm a mac user - it's pretty annoying

Adding another affected case. Max plan subscriber (claude.ai). Cowork sessions (Claude Desktop, Windows) silently ran 1M context during long browser-automation tasks in early June 2026 and billed $54.30 to usage credits — while my included Max plan usage sat at ~2% for the period. Sessions terminated with "API Error: Usage credits required for 1M context."
Support (Fin chatbot, conversations 215474606594428 and 215474606626844) acknowledged that per the model-config documentation, Opus 1M on Max "should be included with your subscription," but stated it cannot retrieve which model variant generated the charges, cannot itemize them by date, and cannot escalate to a human — then cited the non-refund policy. So: charge unidentified, documentation acknowledged to support inclusion, no human review path offered.
Compounding factor: standalone Cowork sessions have no model selector and no way to decline 1M mode (#49649), so this was never a mode I opted into.
Requesting the same review/refund others in this thread received. Happy to provide screenshots and account details to staff.

Opus 4.8 1M context works in the Claude Code CLI but the Windows desktop app silently downgrades it to 200K (Max plan)

Account: Claude Max (20x). Confirmed in the CLI banner: Opus 4.8 (1M context) · Claude Max, and in claude.ai Settings → Usage: Plan usage limits Max (20x).

Summary

On my account, Opus 4.8 with the 1M context window works correctly in the Claude Code CLI but will not run in the Claude for Windows desktop app (Code tab). In the desktop app the 1M model is accepted in the picker yet reverts to standard Opus 4.8 (200K) the moment I submit a prompt, with no error. Because the CLI proves my account is entitled and 1M functions, the defect is isolated to the desktop app surface.

Evidence the account is entitled and 1M works (CLI, Claude Code v2.1.183)

• CLI banner reports the active model as Opus 4.8 (1M context) · Claude Max.
• The /model command in the CLI returned: Set model to Opus 4.8 (1M context) (default) and saved as your default for new sessions.

The defect (Windows desktop app, version 1.14271.0)

• Selecting the 1M model is accepted: the desktop /model command returned Set model to opus[1m] (claude-opus-4-8[1m]).
• The selection holds in the UI and across full app restarts, but reverts to standard Opus 4.8 the instant I submit a prompt. No error or banner appears.
• The 1M variant is also not reliably offered in the desktop model picker, while the CLI offers it.

Ruled out

• Usage exhaustion: usage is far from any limit (Settings → Usage: session 6% used, weekly All models 11% used, Sonnet only 0% used).
• Usage credits: I toggled Usage credits ON and the desktop downgrade still occurs.
• Local configuration: the desktop app does not read settings.json for the model, per #53739 ("Desktop app ignores settings.json 'model' field for new sessions").
• Fable 5: a separate, global suspension (anthropic.com/news/fable-mythos-access), unrelated to this Opus 1M issue.

Documented basis that Max should not be gated

Per How large is the context window on paid Claude plans?:

When using Claude Code with a Pro, Max, Team, or Enterprise plan, Claude Fable 5, Opus 4.8, Opus 4.7, and Opus 4.6 support a 1M token context window.

Pro users need to enable usage credits to access the 1M token context window for Opus models.

The documentation names only Pro for the Opus 1M credit requirement, so a Max account should not be gated. This matches #45390 ("1M context incorrectly requires extra usage on Max plan").

Request

Please fix the Windows desktop Code tab so it offers and honors Opus 4.8 (1M context) for Max accounts, as the CLI already does on the same account, and confirm whether a usage-credits gate is being incorrectly applied to the desktop surface on Max.

Environment

• Claude for Windows desktop app: 1.14271.0
• Claude Code engine: 2.1.183 (per the CLI banner)
• OS: Windows 11