[BUG] [1M] Sonnet 1M context window disappeared after Claude Code update to v2.1.45 — regression on Max plan ($200/mo)

Found 3 possible duplicate issues:

1. https://github.com/anthropics/claude-code/issues/23822
2. https://github.com/anthropics/claude-code/issues/15057
3. https://github.com/anthropics/claude-code/issues/23700

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

I have the same issue. Just so you know, you can still manually select the old 4.5 1m context model using /model claude-sonnet-4-5-20250929[1m] . This is a huge bummer though. I was really looking forward to an improved 1m context model.

A workaround fix I found at the moment is downgrading the version to 2.1.44 and block the auto update, there the 1m model for the x20 max plan still available, and there's 4.6 for both sonnet and opus

I'm experiencing the same issue. After updating to v2.1.45, the 1M context window is no longer available on my Max plan. This is a significant regression — the 1M context was one of the main reasons I subscribed. I hope a fix is rolled out soon.

I successfully worked around the issue for now with:

CLAUDE_CODE_VERSION=2.1.44 claude

I don't know is it relatad but opening extra usage, and adjusting spend limit, shows the models again for me with the latest claude code version.

I'm experiencing the same issue. After updating to v2.1.45, the 1M context window is no longer available on my Max plan. This is a significant regression — the 1M context was one of the main reasons I subscribed. I hope a fix is rolled out soon.

~~I'm using v2.1.50 and still not found sonnet[1m] on the list~~ Now I can use that just set more extra usage.

Try checking if your monthly spending limit under Extra usage in claude web settings is set below how much has been spent so far.

Mine was previously set to 20 max, and I spent 21.13 so far, so i got the same error till I increased it to 50

Looks like 1m context is billed as extra usage now, even on Max plans, which isn't obvious at all.

claude.ai -> Settings -> Usage -> enable Extra usage
After restart, sonnet[1m] works for me again:

Set model to sonnet[1m] (claude-sonnet-4-6[1m]) · Billed as extra usage

The problem keeps happening in newer versions e.g. v2.1.50 ...

and Anthropic support never answer. the customer support for 200 bucks per month is not even real lol

Same exact issue here, even it worked today for me in max plan 200 bucks, v2.1.50 , but out of nowhere still today it just decided to not work ⎿  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_011CYRQB7mvPJp1jNtkZ3naG"}

Same for me as well:

Error: Error during compaction: Error: 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_011CYRS4H9mEureg28ZRBUu7"}

Same for me as well: Error: Error during compaction: Error: 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_011CYRS4H9mEureg28ZRBUu7"}

I ran into the same issue — since the context window has already exceeded 200k, switching back to Opus 4.6 with its 200k context limit means it can't be compacted or rolled back, so I'm essentially forced to abandon most of the work I've already done.

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_011CYRPuhPYmPfW1661ACBwG"}

> Same for me as well: > Error: Error during compaction: Error: 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_011CYRS4H9mEureg28ZRBUu7"} I ran into the same issue — since the context window has already exceeded 200k, switching back to Opus 4.6 with its 200k context limit means it can't be compacted or rolled back, so I'm essentially forced to abandon most of the work I've already done. 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_011CYRPuhPYmPfW1661ACBwG"}

Actually you have a great point , the greed , or i hope just an issue that will be solved, created issues for developers , that can mean delayed work and lost timelines .

UPDATE: BAD NEWS

the workaround is not working anymore, anthropic has removed the Sonnet 1M context window even in the 2.1.44 version. the support never answer, and the capabilities are being reduced in a very drastic way, this is so bad, even if your paying 200 bucks, they are limiting the usage and the regressions are now part of the subscription i guess.

@anthropics

UPDATE: BAD NEWS the workaround is not working anymore, anthropic has removed the Sonnet 1M context window even in the 2.1.44 version. the support never answer, and the capabilities are being reduced in a very drastic way, this is so bad, even if your paying 200 bucks, they are limiting the usage and the regressions are now part of the subscription i guess. @anthropics

This is not fully true, it worked for me in 2.1.50 so its not a version issue

> UPDATE: BAD NEWS > the workaround is not working anymore, anthropic has removed the Sonnet 1M context window even in the 2.1.44 version. the support never answer, and the capabilities are being reduced in a very drastic way, this is so bad, even if your paying 200 bucks, they are limiting the usage and the regressions are now part of the subscription i guess. > @anthropics This is not fully true, it worked for me in 2.1.50 so its not a version issue

how can you check this? i just updated to v2.1.50 and the 1m context is just available as extra usage, but not part of the base plan

> > UPDATE: BAD NEWS > > the workaround is not working anymore, anthropic has removed the Sonnet 1M context window even in the 2.1.44 version. the support never answer, and the capabilities are being reduced in a very drastic way, this is so bad, even if your paying 200 bucks, they are limiting the usage and the regressions are now part of the subscription i guess. > > @anthropics > > > This is not fully true, it worked for me in 2.1.50 so its not a version issue how can you check this? i just updated to v2.1.50 and the 1m context is just available as extra usage, but not part of the base plan

Well i have max plan , 20x , it worked , if i had enabled extra usage option , it didnt actually use the extra money , just kept using the session limit , which was perfect, but all of a sudden it didnt work , but i am sure it did in latest version , i used it 1hr before it stopped working

> > > UPDATE: BAD NEWS > > > the workaround is not working anymore, anthropic has removed the Sonnet 1M context window even in the 2.1.44 version. the support never answer, and the capabilities are being reduced in a very drastic way, this is so bad, even if your paying 200 bucks, they are limiting the usage and the regressions are now part of the subscription i guess. > > > @anthropics > > > > > > This is not fully true, it worked for me in 2.1.50 so its not a version issue > > > how can you check this? i just updated to v2.1.50 and the 1m context is just available as extra usage, but not part of the base plan Well i have max plan , 20x , it worked , if i had enabled extra usage option , it didnt actually use the extra money , just kept using the session limit , which was perfect, but all of a sudden it didnt work , but i am sure it did in latest version , i used it 1hr before it stopped working

is not working for me as expected:

❯ /model                                                                                                                                                                 
  ⎿  Kept model as sonnet[1m]                                                                                                                                            
                                                                                                                                                                         
❯ hola                                                                                                                                                                   
  ⎿  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_011CYRXhv1BGdrE5x8K2oWyn"}                                                                                                        
                                                                                                                                                                         
❯ /model                                                                                                                                                                 
  ⎿  Set model to Default (Opus 4.6 · Most capable for complex work)                                                                                                     
                                                                                                                                                                                                                                                                      
❯ /model                                                                                                                                                                 
  ⎿  Set model to opus[1m] (claude-opus-4-6[1m]) · Billed as extra usage                                                                                                 
                                                                                                                                                                         
❯ hola                                                                                                                                                                   
  ⎿  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_011CYRXmGcFSkjTWU8WyNyxJ"}    

UPDATE: Claude Code v2.1.50, /context shows claude-opus-4-6[1m] · 30k/1000k tokens
with 937K free space. The client clearly recognizes 1M context access.

However, any actual request returns:
"The long context beta is not yet available for this subscription"

Extra usage is enabled. The UI shows 1M. The server rejects it.
This is a server-side gate issue, not a client issue.

❯ /context                                                                                                                                                               
  ⎿  Context Usage                                                                                                                                                       
     ⛁ ⛁ ⛁ ⛁ ⛀ ⛁ ⛀ ⛀ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶   claude-opus-4-6[1m] · 30k/1000k tokens                                                                                    
     ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶    (3%)                                                                                                                     
     ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶   Estimated usage by category                                                                                               
     ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶   ⛁ System prompt: 4.1k tokens (0.4%)                                                                                       
     ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶   ⛁ System tools: 17.5k tokens (1.7%)                                                                                       
     ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶   ⛁ MCP tools: 2.2k tokens (0.2%)                                                                                           
     ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶   ⛁ Memory files: 5.8k tokens (0.6%)
     ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶   ⛁ Skills: 61 tokens (0.0%)
     ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶   ⛁ Messages: 382 tokens (0.0%)                                                                                             
     ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛶ ⛝ ⛝ ⛝ ⛝ ⛝ ⛝ ⛝   ⛶ Free space: 937k (93.7%)                                                                                                
                                               ⛝ Autocompact buffer: 33k tokens                                                                                          
                                                                     (3.3%)                                                                                              
     MCP tools · /mcp                                                                                                                                                    
     └ mcp__filesystem__read_file: 155 tokens                                                                                                                            
     └ mcp__filesystem__read_text_file: 155 tokens                                                                                                                       
     └ mcp__filesystem__read_media_file: 155 tokens
     └ mcp__filesystem__read_multiple_files: 155 tokens
     └ mcp__filesystem__write_file: 155 tokens
     └ mcp__filesystem__edit_file: 155 tokens
     └ mcp__filesystem__create_directory: 155 tokens
     └ mcp__filesystem__list_directory: 155 tokens
     └ mcp__filesystem__list_directory_with_sizes: 155 tokens
     └ mcp__filesystem__directory_tree: 155 tokens
     └ mcp__filesystem__move_file: 155 tokens
     └ mcp__filesystem__search_files: 155 tokens
     └ mcp__filesystem__get_file_info: 155 tokens
     └ mcp__filesystem__list_allowed_directories: 155 tokens

     Memory files · /memory
     └ ~/.../CLAUDE.md: 5.8k tokens

     Skills · /skills

This is a server-side gate issue, not a client issue.

Yes clearly this is the case , Anthropic pulled our access with no warning no nothing , bad move that will for sure upset a lot of people , i mean cmon , frontier models stuck at 200k its small , especially since we are talking about a model which is first of all advertised as 1m

I was also working with it with 2.1.50, just enabling it while reaching the end of the context window to finish the job, then rolling back.
Can't do it anymore after around 3 pm UTC.

It's pretty poor that, despite paying for the most expensive subscription model available, and never hitting any of the limits, we are now not able to use the best model (and not, this is not the best model when it compacts multiple times on every single large/complex session).
Why are we not allowed to use the best models anymore? No one is asking for higher limits, just to use what we're already allotted with the best model and biggest context.
Is Anthropic running out of money/runway?
It's a v poor decision.

I reviewed anthropics/claude-code#26428 and I'm ready to handle it.

• My read: this is documentation-focused work for "Sonnet 1M context window disappeared after Claude Code update to v2.1.45 — regression on M".
• Scope understanding: the critical path is around sonnet, context, where state transitions likely break expected behavior.
• I will focus on these issue-specific points first: sonnet, context, window.
• Distinct insight: this looks like a workflow consistency issue; I will prioritize invariant checks so the fix remains stable under edge cases.
• I can own this end-to-end and keep updates concise.
• Confirm audience and acceptance criteria, then draft reproducible steps.
• Add concrete examples and edge cases so contributors can follow directly.
• I will include targeted validation notes so reviewers can verify quickly.
• Keep progress updates clear (12-24h for first patch, 24-48h for final polish).
• If awarded, I will post first technical findings the same day and deliver an initial patch/validation update within 24 hours.

If this direction works for you, I can start immediately.

Max 20x subscriber here ($200/month). I want to share my experience because I think it perfectly illustrates how badly this was handled.

I used Sonnet 4.5 with 1M context on my Max plan for about three months. It was included. No extra charges. It was genuinely the single biggest reason I justified paying $200/month for Claude Code — being able to work on my entire codebase without auto-compaction destroying my session mid-task was a game changer.

Then the Sonnet 4.6 update dropped, and the 1M context was silently removed from the Max plan. The changelog says "Sonnet 4.5 with 1M context is being removed from the Max plan in favor of our frontier Sonnet 4.6 model, which now has 1M context. Please switch in /model" — as if it's a simple model swap. But it's not. Sonnet 4.6 with 1M now requires extra usage enabled, meaning you pay on top of your subscription. They took a feature I was actively using every single day, removed it without warning, and now want me to pay extra for what I already had.

I even tried downgrading to v2.1.44 to get the old behavior back. The server-side block is already in place — "The long context beta is not yet available for this subscription." So there's no workaround. It's gone.

Let's be honest about the competitive landscape here. Claude is already one of the most expensive AI coding subscriptions on the market. Cursor is $20/month. Copilot is $10/month. Gemini CLI is free. The coding quality gap between Claude and cheaper alternatives has been shrinking fast — Sonnet-tier models are being matched by open-source and competing products. The 1M context window was one of the few things that genuinely set Claude Code apart and made the premium price worth paying.

If even at $200/month you still have to pay extra to access the feature that was your main differentiator, then what exactly are we paying for? Higher rate limits on the same 200k context that everyone else gets? That's not a premium product — that's a volume discount.

Anthropic generated $1 billion from Claude Code in its first six months. The Max plan is clearly profitable. Removing features that paying customers relied on — without announcement, without explanation, without offering any alternative — is not how you treat the users funding your growth.

Bring back 1M context for Max subscribers. Even with reduced rate limits, even with a daily cap. A 1M window with fewer messages is infinitely more valuable than unlimited 200k sessions that get compacted into oblivion every 30 minutes.

I've had access to the 1M context window since Sonnet 4.0, so about 7 months. I can't explain why I was chosen to get that kind of access, but it was amazing! I considered it a privilege granted. When I heard Sonnet and Opus 4.6 would bring in the same 1M context window, but charge extra for it, I guess I understood the early access privilege had run its course and it was time to start paying extra for it when it was needed.

The biggest rub for me is the fact that I no longer have access to any 1M context window in Claude Code. I've enabled Extra Usage. I have $15 in the account and a $200 spend limit per month and I still can't use the 1M context window I've had access to for 7 months.

The salt on the wound is the fact that I'm shown the option to use those models when I run /model but I get an error any time I try and use it. 🫨

API Error: 400 {"message":"The long context beta is not yet available for this subscription."}

I have cancelled the Max Plan. No Point of Paying 200 USD when other vendors are providing 1m Context Window on anthropic models at a similar discounted price. Claude Code became unusable for us.

I have the same issue. Just so you know, you can still manually select the old 4.5 1m context model using /model claude-sonnet-4-5-20250929[1m] . This is a huge bummer though. I was really looking forward to an improved 1m context model.

I'm not sure how I missed this 🤦‍♂️ ...but this worked for me! Thank you!!

Same issue here — 4+ months of Sonnet 4.5 1M beta access, now completely broken
I've been a Max subscriber actively using sonnet[1m] for over 4 months with zero issues. This was a core part of my daily workflow — not an experimental feature I was "trying out."
After the Sonnet 4.6 / v2.1.45 update, both sonnet[1m] and opus[1m] are now returning "not available for your account."
The real damage: I have multiple active sessions with ~400K tokens of context built up over ongoing development work. These sessions cannot be meaningfully summarized or recreated — the context continuity IS the value. Losing 1M access doesn't just block new work, it destroys work already in progress.
What makes this especially frustrating:

Sonnet 1M was stable for 4+ months — this is a regression, not a missing feature
The docs still don't list Sonnet 1M as restricted for subscribers (only Opus 1M)
No communication whatsoever — no email, no changelog note, nothing
Customer support gives generic non-answers with no escalation path

If this is intentional, Anthropic owes long-term beta users a migration path and clear communication. If it's a bug, please prioritize — active work sessions are being lost every day this remains broken.

Same issue here — 4+ months of Sonnet 4.5 1M beta access, now completely broken I've been a Max subscriber actively using sonnet[1m] for over 4 months with zero issues. This was a core part of my daily workflow — not an experimental feature I was "trying out." After the Sonnet 4.6 / v2.1.45 update, both sonnet[1m] and opus[1m] are now returning "not available for your account." The real damage: I have multiple active sessions with ~400K tokens of context built up over ongoing development work. These sessions cannot be meaningfully summarized or recreated — the context continuity IS the value. Losing 1M access doesn't just block new work, it destroys work already in progress. What makes this especially frustrating:

Have you tried opening those active sessions up, and running /model claude-sonnet-4-5-20250929[1m]? That got me the access to 1M Sonnet 4.5 again...You might be able to get back to work on those large running sessions!

NOTE: I think it's billed as extra usage now though, so no longer part of the $200 Max plan...because I now see:

/model claude-sonnet-4-5-20250929[1m] 
  ⎿  Set model to claude-sonnet-4-5-20250929[1m] · Billed as extra usage

Just started happening for me as well!

Confirming this is still broken as of Claude Code v2.1.50 (Feb 23, 2026).

Setup:

• Claude Max subscription
• Claude Code v2.1.50
• Default model (Opus 4.6) works fine

Repro:

1. /model → select sonnet → claude-sonnet-4-6[1m]
2. Send any message
3. Get: API Error: 400 — "The long context beta is not yet available for this subscription."
4. /model → select opus → claude-opus-4-6[1m]
5. Same error.
6. /model → select Default (Opus 4.6) → works fine.

Both explicit 1M model selections fail. Only the default model route works. This has been broken since at least v2.1.45 and is still not resolved. Max plan users are paying for capabilities they can't access through the model switcher.

> 저도 같은 문제네요. Sonnet 4.5 1M 베타 액세스를 4개월 넘게 사용했는데, 이제 완전히 사용할 수 없게 됐습니다. Max 구독자로서 4개월 넘게 sonnet[1m]을 문제없이 사용해 왔습니다. 이건 제 일상적인 작업 흐름의 핵심이었고, 단순히 "시험 삼아 사용해 보는" 실험적인 기능이 아니었습니다. Sonnet 4.6 / v2.1.45 업데이트 이후 sonnet[1m]과 opus[1m] 모두 "계정에서 사용할 수 없습니다"라는 메시지가 표시됩니다. 진짜 문제는 제가 진행 중인 개발 작업으로 축적된 약 40만 개의 컨텍스트 토큰이 있는 여러 활성 세션이 있다는 것입니다. 이러한 세션은 의미 있게 요약하거나 다시 만들 수 없습니다. 컨텍스트의 연속성이 바로 핵심 가치입니다. 1M 액세스 권한을 잃으면 새로운 작업이 중단될 뿐만 아니라 이미 진행 중인 작업도 망가집니다. 특히 답답한 점은 다음과 같습니다. 활성화된 세션을 다시 열고 실행해 보셨나요 /model claude-sonnet-4-5-20250929[1m]? 저는 그렇게 해서 1M Sonnet 4.5에 다시 접근할 수 있었어요. 그러면 대용량 실행 세션도 다시 작업할 수 있을지도 몰라요! 참고: 제 생각에는 이제 추가 사용량으로 청구되는 것 같아서 더 이상 200달러 최대 요금제에 포함되지 않는 것 같습니다... 왜냐하면 이제 다음과 같이 표시되기 때문입니다. `` /model claude-sonnet-4-5-20250929[1m] ⎿ Set model to claude-sonnet-4-5-20250929[1m] · Billed as extra usage ``

Thanks for the tip — can confirm /model claude-sonnet-4-5-20250929[1m] works for me too! My active sessions with ~400K context are alive again.
And yes, I'm also seeing "Billed as extra usage" now. I have $50 in extra usage credits from the Opus 4.6 launch promo, and so far the balance hasn't changed after several exchanges. So it's unclear whether it's actually billing extra or if the message is just a UI indicator that the 1M feature has moved to the extra usage tier.
Also worth noting: opus[1m] is working for me now too under the same "extra usage" billing. I had to toggle my extra usage setting off and on in account settings — didn't work immediately, but about 10 minutes later everything unlocked across all my machines.
TL;DR for others hitting this issue:

Make sure extra usage is enabled in your account settings
If it's already on, try toggling it off → on
Wait ~10 minutes for backend propagation
Use /model claude-sonnet-4-5-20250929[1m] for Sonnet 4.5 1M access
opus[1m] may also work under extra usage billing

Still frustrating that none of this was communicated. 4+ months of stable Sonnet 1M access silently moved to extra usage billing with zero notice.

Hey folks! There are two matters being discussed in this thread.

1. Today, there was a bug (which we've since fixed) that led to the API responding with a 400 error saying "The long context beta is not yet available for this subscription". (See #27950).
2. As of Sonnet 4.6 and Opus 4.6, the 1M context modes are indeed charged as extra usage. To set /model opus-4-6[1m] (for example) you'll need to have purchased extra usage credits.

What about the Sonnet 4.5 1M context?

@cwilso03 Here's what Claude Code says:

❯ /model sonnet-4-5[1m]                                                                                                                  
  ⎿  Model 'sonnet-4-5[1m]' not found                                                                                                    
                                                                                                                                         
❯ /model sonnet-4-6[1m]                                                                                                                  
  ⎿  Sonnet 4.6 with 1M context is not available for your account. Learn more:                                                           
     https://code.claude.com/docs/en/model-config#extended-context-with-1m                                                               

It seems that the behavior shown above is intended, and they will not restore model sonnet-4-5[1m].

Yeah, I see that. Would be nice if they at least let us continue using Sonnet 4.5 1M as part of the subscription plan, since we've had access to it for a while (speaking as a Max 20x subscriber). Like others mentioned in this thread, I was in a session that had over the 4.6 standard 200,000 tokens in it when they shut off the access with no advance notice this morning. Not a great customer experience, to be sure.

workaround in 2.1.50: /model claude-sonnet-4-5-20250929[1m]

It says it will be billed as extra usage, but I don't have extra usage enabled on my account, so clearly that's not the case. 20x subscribers still get 4.5[1m] As far as I can tell.

Same issue here — 4+ months of Sonnet 4.5 1M beta access, now completely broken I've been a Max subscriber actively using sonnet[1m] for over 4 months with zero issues. This was a core part of my daily workflow — not an experimental feature I was "trying out." After the Sonnet 4.6 / v2.1.45 update, both sonnet[1m] and opus[1m] are now returning "not available for your account." The real damage: I have multiple active sessions with ~400K tokens of context built up over ongoing development work. These sessions cannot be meaningfully summarized or recreated — the context continuity IS the value. Losing 1M access doesn't just block new work, it destroys work already in progress. What makes this especially frustrating: Sonnet 1M was stable for 4+ months — this is a regression, not a missing feature The docs still don't list Sonnet 1M as restricted for subscribers (only Opus 1M) No communication whatsoever — no email, no changelog note, nothing Customer support gives generic non-answers with no escalation path If this is intentional, Anthropic owes long-term beta users a migration path and clear communication. If it's a bug, please prioritize — active work sessions are being lost every day this remains broken.

Couldn't have said it better myself, this is the core strength in my current codebase, I now cannot repeat my workflow without making huge changes to how I manage the context. I will cancel if nothing comes of this, this is uncalled for and extremely unfair after having complete access for 6 Months with my Claude Max plan that im paying $390 CAD for a month.

@mcgogo64-cell try /model claude-sonnet-4-5-20250929[1m]

Related issue: after v2.1.51 update, 1M context models were silently moved to Extra Usage-only billing. The model picker now shows "Billed as extra usage" but the changelog only mentioned "model picker improvements." No billing change was documented. See #28927 for JSONL evidence showing identical usage billed differently before/after the update.

just updated to v2.1.61 they now officially removed the 1m context models in /model list. im now getting rate limit error and cant even continue my existing conversations because they exceed 200k context because i have been using 1m context more than a week now. well FCK Y** ANTHROPIC.

The 1m model still not available, i updated the account with extra usage, but it was charged, the problem keeps happening, and is a big downgrade an a regression

Max 20x subscriber — Two ongoing issues

Environment: Windows 11, Claude Code latest, Max 20x ($200/mo)

1. 1M context should be included in Max 20x subscription

I'm a Max 20x subscriber paying $200/month. It's hard to understand why 1M context requires extra usage billing on top of the highest-tier subscription.

The current policy feels inconsistent — Max 20x is marketed as the premium tier, yet the flagship 1M context feature requires additional payment. This creates a frustrating experience for users who chose the most expensive plan expecting full access to all features.

2. Weekly usage jumping 7-8% unexpectedly

Separately, I'm experiencing an issue where my weekly usage percentage suddenly spikes by 7-8% without corresponding activity. This has been happening consistently in recent weeks (as of March 2026). There seems to be background token consumption that isn't transparently reported.

Both issues point to a broader concern about billing transparency and value alignment for Max 20x subscribers.

The extra usage only kicks in if you utilize that 1M context amount. I used the Opus 4.6 1M for several weeks without costing me any extra usage simply because I never needed that full 1M context.

I had to have Claude explain to me how it works...

In Claude Code v2.1.75:

Opus now defaults to 1M context · 5× more room, same pricing

cc @iceepc-netizen @BHCC2025 @mafiaboyhacker @juanarangot @KJ7LNW

In Claude Code v2.1.75: > Opus now defaults to 1M context · 5× more room, same pricing

What does same pricing mean? Is this included as part of the subscription? Or do you still have to pay per token for the one million model?

@KJ7LNW It's the new default, as of today:

Claude Opus 4.6 and Sonnet 4.6 have a 1M-token context window.

See here: https://claude.com/blog/1m-context-ga

1M context is now included in Claude Code for Max, Team, and Enterprise users with Opus 4.6. Opus 4.6 sessions can use the full 1M context window automatically, meaning fewer compactions and more of the conversation kept intact. 1M context previously required extra usage.

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.