It started simple. Use the bot. It’s free. Or pay twenty bucks. Get more power. A clean line between consumer and power user.
Anthropic is smearing that line.
Starting July 12 at 23:59 PT the math gets messy. If you’re on the $20. Or the $100. Or the big boy $200 plan you’re paying extra to use Claude Fable 5. The consumer face of their heavyweight Mythos 5 model. This is a first. A frontier lab gating top-tier intelligence behind a pay-per-use meter. Never seen it before.
The rates? Developer API rates. No consumer discount.
– Send 1 million tokens: $10
– Get 1 million tokens back: $50
Do the math. You send a million words to the bot. It sends a million back. That’s $60. On top of your subscription. So your “budget” $20 plan costs you $80 for the month. Roughly the price of five months of Amazon Prime just to talk to a chatbot.
Is that a lot?
A million tokens equals 750,00 words. Longer than all the Lord of the Rings books combined. But power users don’t hit those numbers by reading Tolkien. They hit them by forcing models through complex reasoning chains. Hidden processing. The kind that burns tokens faster than a house fire.
For years AI labs hated variable billing. They loved flat subscriptions. Predictable revenue. Control over who got access.
“Having an unlimited [AI] plan… is like having an unlimited electricity plan. It just doesn’t make sense.”
— Nick Turley, formerly of OpenAI
The shift isn’t sudden. Cursor cut unlimited coding access. Anthropic already charges big business by usage. Maybe they’re cleaning up their balance sheet for an IPO. Or maybe they just realized infinite intelligence is expensive to serve.
Anthropic won’t banish subscriptions entirely. Not yet.
Spokesperson Reem Ateyeh tells WIRED they want Fable 5 back in the plan tiers “when sufficient capacity allows.” Capacity being the bottleneck. They’ve bought billions in cloud space from Amazon and SpaceX. Still not enough. Ever will there be enough? Unlikely.
So why charge?
A test. A market stress test for consumer AI appetite.
Anthropic has played enterprise second fiddle to ChatGPT’s cultural dominance. Until recently. Now Claude hits 245 million users a month. Double its size from February. Still nowhere near ChatGPT’s 1.1 billion or Gemini’s 662. But the growth rate is violent.
The backlash is equally loud. Critics say Anthropic is gouging users while harvesting the world’s data.
The company ignores the noise. They bet on being the Apple of AI. There will always be people who pay a premium. For quality. For privacy. For the specific texture of a well-reasoned answer.
“People aren’t asking ‘Do I need the best?’ They are asking ‘Am I the person who deserves the best?'”
I talk to VCs. Fund managers. Engineers. They pay up. They want the edge. Right now that edge is Claude. OpenAI has its GPT models. Google has Gemini. But in certain circles the badge is Claude.
Anthropic’s stance? No ads. Ever.
Competitors like Google will flood free tiers with promotions. OpenAI might too. Anthropic chose hard pricing over soft annoyance.
The golden era is over. Subsidies are gone. You want the smartest bot in the room?
Write a check.
