[BUG] Opus 4.6 with [1m] context window returns "long context beta not available" on Max subscription
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
Attempting to use Claude Opus 4.6 with the 1M context window via /model opus[1m] or /model claude-opus-4-6[1m] returns:
API Error: 400 {"type":"error","error":{"type":"invalid_request_error","message":"The long context beta is not yet available for this subscription."},"request_id":"req_011CXqWqt1M7ZkA4iBcYGDsN"}
This is on a Max subscription. The Opus 4.6 announcement states "Opus 4.6 features a 1M token context window in beta" and the API docs confirm Opus 4.6 supports 1M context via the context-1m-2025-08-07 beta header.
What Should Happen?
Max subscribers should be able to use opus[1m] or claude-opus-4-6[1m] in Claude Code, consistent with the model's advertised capabilities.
Error Messages/Logs
Steps to Reproduce
Log in to Claude Code with Max subscription (/login)
/model opus[1m] (or /model claude-opus-4-6[1m])
Send any message
Error returned on first message
Claude Model
Opus
Is this a regression?
Yes, this worked in a previous version
Last Working Version
_No response_
Claude Code Version
2.1.32
Platform
Anthropic API
Operating System
Ubuntu/Debian Linux
Terminal/Shell
Terminal.app (macOS)
Additional Information
_No response_
29 Comments
Found 3 possible duplicate issues:
This issue will be automatically closed as a duplicate in 3 days.
🤖 Generated with Claude Code
Same:
/model opus[1m]
API Error: 400 {"type":"error","error":{"type":"invalid_request_error","message":"The long context beta is not yet available for this subscription."},"request_id":"req_011CXqcQZGUZMEZKxosRVajr"}
This is unfair for $200 usd a month, we have been requesting it for +4 months now.
It's a feature, not a bug.
Opus 4.6 [1M] context is not available at launch for Pro, Max, Teams, or Enterprise subscribers.
https://code.claude.com/docs/en/model-config#extended-context-with-1m
<img width="1672" height="960" alt="Image" src="https://github.com/user-attachments/assets/498b43cd-bfe1-4b7b-b833-cb78a4bc51e2" />
It's not just for Opus, and it's certainly not a feature:
Their suppor system says:
They don't clarify who qualifies for this access, and why we don’t currently have it. And this has been going on for more than 4 months now.
Then this is very confusing and dissapointing, because no one knows what "in beta" means. It's all good that there some improvements in the model, but the real improvement most of us want is the context size. The 4.5 was already good if not for the context size.
<img width="633" height="134" alt="Image" src="https://github.com/user-attachments/assets/0103fd46-dfa3-4d1f-8647-7dfff1a47396" />
This appears to be the correct answer, whether or not we like it. Communication could have been clearer.
okay, so I just had this happen to me - Lunchtime letdown for my Max subscription after the press release and drop, after I burned my weekly cap yesterday before being able to try it. Nice big refactor planned, going to slum it with 200k.
I realize now, I read that first period, as a comma in the image below, and because I thought I read what I wanted to see (Max has access) and did not read it 'fully' .. it got me here..psyched up and disappointed. This should have been made clearer, and I think Anthropic knows it with their extra usage promo for Max subscribers that was updated today.
<img width="1568" height="233" alt="Image" src="https://github.com/user-attachments/assets/0be8b887-4c4c-4f96-beb6-44a822b5b5f0" />
Found the fix https://netstatz.com/enable-claude-opus-4-6-1m-context/
Nice. Obviously they don't want max subscribers to have 1m so presumably they will close this workaround though?
Otherwise they'd surely just make it work...
Yep.. just started borking. Damm it. Does not allow selection in new sessions, and current 1M either fail with api error at session start (if loading with 1M active) and existing ones context ends right away, even with 1000k in my statusline. Boo
For a small while, I thought the API signal might have been intentional .. I know 20x users that have it, so it is rolling out to some Max Subscribers.
Finally! The 1M models are available now!!!
<img width="1195" height="231" alt="Image" src="https://github.com/user-attachments/assets/733fca1d-b1c9-49ef-a112-7990410ed748" />
I haven’t gained access yet, though it’s great that you already have it. 200k tokens is not enought,
Anthropicans reading this: your highest-paying individual subscribers are the ones most likely to use your most capable model's most capable feature. Charging them extra for it doesn't just leave money on the table, it actively discourages upgrades.
I was just about to upgrade from Pro to Max. My primary use case is Claude Code with Opus 4.6[1m]. Here's what I discovered: the 1M context window (a headline feature of Opus 4.6) requires Extra Usage on every subscription tier, including Max 20x. Pro and Max subscribers hit the exact same paywall. The rational move for someone with my use case is to stay on Pro at $20/month and fund Extra Usage directly, rather than pay $200/month for Max and still fund Extra Usage on top.
That's $180/month in potential upgrade revenue Anthropic isn't capturing (per user, per month) from exactly the power-user segment most likely to upgrade.
As founding president emeritus of the Boston Product Management Association, I spend a lot of time thinking about pricing tiers and upgrade incentives. And I'd genuinely like to see Anthropic succeed. The fundamental rule is: each tier must unlock meaningfully more value to justify the price increase. For the fastest-growing segment of power users, those building with Opus[1m] in Claude Code, the upgrade ladder is broken.
In contrast, Google offers Gemini CLI with their 1M-context Pro model free to anyone with a personal Google account (60 requests/min, 1,000 requests/day, no credit card required). The model quality debate is fair game, but the perception gap is devastating in developer communities. "Free with 1M context" sitting next to "$200/month plus Extra Usage for 1M context" writes its own narrative.
Dario, if you value this product consulting, I'd welcome a lifetime Max subscription :-)
Indeed. Given how slow Claude has been for me today I'm a bit concerned about this recently added "pay more to make it go faster" option that popped up on my terminal the other day.
I hope those of us paying a decent chunk of cash each month for this aren't going to be left in the slow lane because we're not paying a decent chunk of cash plus another decent chunk of cash...
What the hell is going on! I lost my 1M models suddenly. It was working fine and then I started getting errors that I reached my Rate Limits. I restarted Claude Code and now 1M models are gone. And this is considering that today was my renewal day for the subscription. So, they took the money in the morning and disabled 1M models in the evening.
This is getting ridicules
<img width="1183" height="340" alt="Image" src="https://github.com/user-attachments/assets/5c2230c5-ebf5-4edc-b17a-5091d53412d3" />
Additional repro: Same account, different auth methods, different results
/login→ claude.ai auth):opus[1m]andsonnet[1m]both return "Rate limit reached" immediately/login→ API account): Bothopus[1m]andsonnet[1m]work fineThis confirms the issue is in the subscription auth path, not in account billing or credits. The API backend accepts
[1m]requests when authenticated via API key but rejects the identical request when authenticated via Max subscription - on the same account.sonnet[1m]worked intermittently earlier today via subscription auth, then stopped.opus[1m]never worked via subscription auth.Update: Resolved - and a detailed UX/documentation improvement proposal
Resolution
The fix for my case: Enable "Extra usage" in Settings > Usage on claude.ai, with a payment method attached. The
[1m]context models require extra usage billing even on Max plan. Once enabled, bothopus[1m]andsonnet[1m]work.The debugging journey (why this took over an hour to figure out)
Here's the full sequence of confusion I went through as a Max subscriber:
opus[1m]→ instant "API Error: Rate limit reached"/statusshows plenty of session and weekly quota remainingsonnet[1m]→ worked briefly, then also started failing[1m]→ thought "I have a Max plan, that should cover it"/loginwith API key →[1m]works./loginwith Max subscription →[1m]fails. Same email.[1m]fitsWhat the current documentation says (and where it falls short)
What exists:
The model configuration docs do say:
The extra usage article explains what extra usage is and how to enable it.
What's missing:
"Extra usage is required for long context requests."(visible in error logs from issues #27179 and #26496) but Claude Code surfaces it as "Rate limit reached."opus[1m]and see "Billed as extra usage," there's no link to enable extra usage. If extra usage isn't enabled, there's no indication that's why it will fail.[1m]- otherwise you get a generic error with no guidance.---
Proposed improvements
1. Fix the error message (5-minute fix, biggest impact)
Current:
Proposed:
The API already returns the correct message (
"Extra usage is required for long context requests."). Claude Code just needs to surface it instead of genericizing it to "Rate limit reached."2. Add an inline warning when selecting
[1m]in the model pickerWhen a user selects
opus[1m]orsonnet[1m]via/model:Current:
Proposed (when extra usage is NOT enabled):
Proposed (when extra usage IS enabled):
3. Create a dedicated "1M Context Window" support article
Suggested URL:
support.claude.com/en/articles/1m-context-window-billingContents:
4. Add a "1M Context" section to the existing "Using Claude Code with your Pro or Max plan" article
The current article at support.claude.com/en/articles/11145838 doesn't mention 1M context at all. Add a section:
---
Summary
The information technically exists across 3-4 different pages, but:
The #1 fix is the error message. Everything else is documentation polish. That single string change would have saved me (and likely hundreds of others in these threads) an hour of debugging.
I have had access to [1m] for Sonnet and Opus until this morning on Max plan with extra usage enabled. Today I got the API error message.
same here everything was going great on a opus 4.6 (1m) logged in with Max plan and then all of a sudden I started getting the error and had to manually edit the session json down so I could get it to compact in opus 4.6 200k
This makes sense. Thanks for explaining this.
Now I realized, that I did have extra usage enabled with the $70 Anthropic was giving as a gift and this is why I had the 1M models available. As soon as I used up that amount I started getting the rate related error.
I had it enabled, even added funds, had plenty of funds, was 506k of 1000K context opus 4.6 (1m); all it would do was display the error. /compact would not work just kept displaying the error. Max plan.
I use max20x and this is very disappointing to me. Until recently, I was very happy about Anthropic and was proud to be a Claude user. Now, I will be considering other options. :(
So is it Anthropic's position that 20x Max users will not have access to 1m in context all; without paying for extra usage; and they are trying to onboard people in paying for extra usage, who already shell out 200/mo, by giving a pittance of extra usage in coupons to hook people?
Again, I share the sentiment of the majority of Max users who think it's insulting, goofy and an active detriment to the consumer base of Anthropic's tooling to not give the flagship subscriptions the ability to access the 1m tokens without shelling out more cash.
While you're in "beta" you want to limit the amount to prevent abuse, fine; even after beta you want to say max 200k+ token session usages a week is x10-20, fine; but to outright make folks pay more, when they are already paying $200/mo is more than just questionable, when Google and Codex have higher context availability. We already have a token limitation, just because it's in one session vs multiple sessions, it's still a token limitation. This reads as just greedy, or not being able to keep up with the competition.
What's the scoop here, am I misreading the room? I would love to be wrong. Is there somewhere to read on Anthropic's intentions for this? Is this going to be how it is moving forward? The extra usage requirement to utilize one of the most promising features the current models have to offer?
Yes, they need to reconsider this and considering the other big bit of news
that was coming out recently I don’t seem to be making very good strategic
decisions at the high-level. If they want to add one more tear and call it
250 at least then it’s an agreement that Maintained so that trust is
maintained. The biggest issue here is the erosion of confidence and that
confidence is what makes people consider long-term relationships.
--------
Aaron Lippold
@.***
260-255-4779
twitter/aim/yahoo,etc.
'aaronlippold'
On Fri, Feb 27, 2026 at 18:38 Zech Walden @.***> wrote:
Man voice to text sucks on an iPhone lol
--------
Aaron Lippold
@.***
260-255-4779
twitter/aim/yahoo,etc.
'aaronlippold'
On Sat, Feb 28, 2026 at 13:04 Aaron Lippold @.***> wrote:
Rejoice: https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md#2175
I'm actually still not seeing the 1m context window, just 200k. I had extra usage off, which I just turned off, then logged in again still didn't see it. Disabled my oauth token, logged in, still didn't see it. Waiting to see if the new 5-hour session might kick it on...
EDIT: Still not seeing it after moving onto the new 5-hour window. I'm on the Max 20x plan.
EDIT 2: Ok, I'm seeing it now. I had to run
claude updateinside of claude code evening though I was already running Claude Code v2.1.76, which was the up-to-date version of claude code at the time...EDIT 3: Added back my oauth token, no issues.
It has been working for me for the last two days
<img width="653" height="41" alt="Image" src="https://github.com/user-attachments/assets/24a236b9-f38f-4d90-bb03-ba3a03972b90" />
This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.