"This feels like I've been bait and switched TWICE now." — Claude Max subscriber, January 2026
What Anthropic Did:
- December 25-31, 2025: Doubled usage limits as a "holiday gift"
- January 1, 2026: Reverted to "normal" limits
- January 6+, 2026: Limits became MORE restrictive than ever before
What Users Got:
- 60% reduction in available tokens (per user analysis)
- Sonnet limits hit in 10-15 minutes
- Opus unavailable after session start
- 10x variance in burn rates (5.6%/hr to 59.9%/hr on the SAME account)
- Criticism allegedly censored on Discord
Anthropic's Response: "The concerns appear to be largely a response to the resumption of normal limits."
User Response: Trustpilot rating: 1.5/5 stars (646 reviews). Words used: "scam," "bait and switch," "theft."
Anthropic announced a holiday promotion:
"We're giving Pro and Max subscribers a holiday gift—2x usage limits from December 25 through December 31. During this period, your plan's usage limits will be doubled, giving you more room to explore Claude's capabilities." — Claude Help Center
Users were delighted. Productivity soared. Developers integrated Claude deeper into their workflows.
The bonus ended. But users noticed something wrong — limits weren't just back to "normal." They were worse.
"Claude Code users are reporting that they are hitting usage limits more quickly, even during light tasks." — dev.ua, January 6, 2026
The Register broke the story: a wave of complaints flooded Claude Developers Discord and Reddit. One customer shared analysis showing ~60% reduction in token usage limits.
"Developers say that interactions with models have become 'costing' more tokens, which is why Pro/Max limits are being reached faster: posts include scenarios where Sonnet limits are triggered in 10-15 minutes, and Opus becomes unavailable for some users shortly after the start of a session." — The Register
GitHub Issue #17084 documented the escalation:
"Since the weekly limit reset on Thursday, January 8, 2026, Opus 4.5 usage limits have been significantly reduced compared to all previous weeks since Opus 4.5 launched in November 2025."
"After just one day (Friday, January 9), I consumed over 27% of my weekly 'All models' (Opus 4.5) limit."
"This appears to be a new, more aggressive limit reduction beyond what was reported after January 1st."
A Max 20x ($200/month) subscriber instrumented their API calls and discovered damning evidence:
| Session | Duration | Burn Rate | Status |
|---|---|---|---|
| Jan 30, 18:41 - Jan 31, 02:37 | 7.9h | 9.3%/hr | Normal |
| Jan 31, 20:28 - Jan 31, 22:01 | 1.5h | 56.0%/hr | 🔴 6x faster |
| Feb 01, 15:00 - Feb 01, 16:40 | 1.7h | 59.9%/hr | 🔴 6x faster |
Same user. Same plan. Same workload. 10x variance.
The user's conclusion:
"If limits simply 'returned to normal,' they would return to a consistent normal rate. The 10x variance between sessions is either:
- A bug in quota accounting (which Anthropic should acknowledge and fix)
- An intentional server-side change (which should be disclosed per ToS)"
This issue is titled: "Legal liability claim ready."
| Issue | Title | Upvotes |
|---|---|---|
| #16157 | Instantly hitting usage limits with Max subscription | 490+ |
| #16270 | Usage limits bugged after double limits expired | 43 |
| #16856 | 4x+ faster token consumption than previous versions | — |
| #16868 | Credit consumption rate increased 3-5x after Jan 1st | — |
| #17084 | Opus 4.5 limits most restrictive since launch | — |
| #22435 | Inconsistent quota accounting - Legal claim ready | — |
| #22439 | Limit reached despite 85% quota remaining | — |
"This is a total scam" — I can't have a ten minute conversation with my paid subscription without hitting limits, and unfortunately have paid for a year of this misery and see no way to seek a refund.
"This product is the definition of 'bait and switch.'" — The AI itself is great, but the company behind it (Anthropic) has implemented limits that make it impossible to use seriously.
"Claude are scammers" — Even with only 1 person and a €240 paid subscription, Claude limits your work after only 1-2 hours work and you have to wait another 4 hours. I am now searching for an alternative in China without limitations like DeepSeek.
"SCAM!" — I just paid 18 bucks to remove the 4 hour wait, then I talked to Claude for like 15 minutes and then it blatantly told me that I have to upgrade or wait another 5 Hours.
"4 messages costed me approximately 10% of my $200/mo max plan weekly limit." — Mr. Amodei I love your tool, but please do not become another money grabbing corporation.
"This feels like I've been bait and switched TWICE now."
"Whether you call it bait and switch, predatory pricing, or consumer deception, it's just plain unfair."
"I'm literally paying for a service I can't use."
"$100 Max is the new PRO — PRO was just a trial."
"Do I understand correctly that people (and 'enterprises') sign contracts that specify how much they pay, but not what they are getting for it? Neither the users nor the vendor seem to be able to specify what a 'token' is."
"When a service provider refuses to provide an agreed upon service, I stop using it."
Multiple users suspect a deliberate strategy:
"Anthropic bumped limits over the holidays to drive more usage. Which combined with Opus's higher token usage and weird oddities in usage reporting/capping makes me suspect they want to drive people from Pro → Max without admitting it." — HN Comment, Jan 30, 2026
The pattern:
- Give users a taste of higher limits during holidays
- Make them dependent on that capacity
- Revert to limits that feel unbearable by comparison
- Users upgrade to Max to get what they had for free
The evidence:
- Max accounts now have "the former Pro limits"
- Pro accounts are "even worse than the free tier"
- Users report: "You will hit limits in just a few prompts, making the service unusable"
Users claim their criticism was silenced:
"Discussions have accused Discord moderators of deleting criticism and banning, but Anthropic denies censorship and says that the blocks are related to violations of community rules." — The Register
"No response from Anthropic on related issues despite 237+ upvotes." — GitHub Issue #17084
One GitHub issue (#11810) compiled complaints and noted:
"Priority: CRITICAL — Directly impacting subscription retention and revenue"
The response? Silence.
Some users speculate the timing isn't coincidental:
"He speculates that the changes represent an attempt to reduce costs prior to Anthropic's expected public stock offering, a claim Anthropic flatly denies." — The Register
The logic:
- IPO requires showing profitability or path to it
- AI inference is expensive (~$15-75 per million tokens for Opus)
- Reducing actual usage while maintaining subscription revenue = better margins
- Holiday bonus = temporary loss leader to hook users before tightening
Anthropic denied this connection.
| Plan | Price | Advertised |
|---|---|---|
| Pro | $20/mo | 5x Free usage |
| Max 5x | $100/mo | 5x Pro usage |
| Max 20x | $200/mo | 20x Pro usage |
According to community analysis:
- 60% reduction in available tokens post-holiday
- 3-5x faster credit consumption
- 10x variance in burn rates within same account
- Limits hit in 10-15 minutes on Sonnet
- Opus unavailable shortly after session start
If you're paying $200/month for Max 20x and getting:
- 60% less tokens than expected
- You're effectively paying $500/month equivalent for what you receive
- Or: Anthropic owes you $120/month in unreceived service
Users are voting with their wallets:
"I am now searching for an alternative in China without limitations like DeepSeek." — Trustpilot review
"As of now I'm using Google's Antigravity, which is inferior yes, however the limits are ABSURDLY bigger." — Trustpilot review
DeepSeek pricing:
- Input: $0.27/M tokens
- Output: $1.10/M tokens
- No artificial usage caps
Claude Opus pricing:
- Input: $15/M tokens
- Output: $75/M tokens
- Plus artificial usage caps on top
DeepSeek is 30-50x cheaper AND doesn't cut you off mid-task.
From The Register:
"The concerns appear to be largely a response to the resumption of normal limits."
From the Help Center:
"From December 25 to December 31, 2025, Anthropic temporarily doubled the limits as a 'holiday gift' and to load free computing resources, and now the limits have returned to the standard."
Translation: "We're not doing anything wrong. You just got used to something better."
The Problem: Users with instrumented data show limits are worse than pre-holiday, not just "returned to standard."
As if token issues weren't enough, January 2026 brought a billing catastrophe:
"SEVERE BILLING ERROR — On January 12, 2026, Anthropic's billing system charged me €1,630.98 for DUPLICATE subscriptions... TOTAL UNAUTHORIZED: €3,221.37 for what should be €22.14 monthly subscription." — Trustpilot review
Anthropic's status page acknowledged:
"We are investigating reports that some new customer subscriptions are charging customers without properly granting subscription access, which is also leading to accidental double or triple payments."
So users got: fewer tokens AND overcharged for them.
-
Why do burn rates vary 10x for the same user, same plan, same workload?
-
Why are January 2026 limits MORE restrictive than November 2025 (pre-holiday)?
-
Why do dashboards show available quota when Claude blocks usage?
-
Why is there no response to GitHub issues with 490+ upvotes?
-
What is the actual token allocation for each plan? (Not "5x more" — actual numbers.)
-
Why were Discord criticisms deleted if there's nothing to hide?
| Claim | Reality |
|---|---|
| "Holiday gift" | Bait to increase dependency |
| "Returned to normal" | Worse than pre-holiday |
| "5x/20x more usage" | Undefined, variable, inconsistent |
| "Premium experience" | Hit limits in 10-15 minutes |
| Trustpilot rating | 1.5/5 stars |
The Pattern:
- ✅ Give generous holiday bonus
- ✅ Users integrate Claude deeper into workflows
- ✅ Remove bonus + reduce limits further
- ✅ Users forced to upgrade or leave
- ✅ Silence criticism
- ✅ Blame users for "misunderstanding"
- The Register: Claude devs complain about surprise usage limits
- dev.ua: Claude Code users reported a drop in limits
- Techzine: Claude developers concerned about usage limits
- #16157 — Instantly hitting limits (490+ upvotes)
- #17084 — Most restrictive since Opus launch
- #22435 — 10x burn rate variance, legal claim ready
- #11810 — "Bait and Switch" compilation
- Trustpilot: Claude.ai — 1.5/5 stars, 646 reviews
- Holiday 2025 Usage Promotion — Claude Help Center
Data collected: February 3, 2026
P.S. If Anthropic reads this: Your users aren't stupid. They have logs. They have data. And they're comparing notes.