Source linked

Anthropic's Fable Model Goes API-Only June 22 - The End of Unlimited AI

The safe version of Anthropic's latest model, Fable, will disappear from subscription plans in under two weeks, marking a shift to per-token pricing that signals the death of the unlimited AI subscription model.

anthropicfableclaudeai subscription modelinference costsapi pricing

Anthropic just gave Fable, its latest frontier model, a 12-day shelf life on subscription plans. After June 22, you'll pay per token.

Fable is the publicly safe variant of Mythos, the model Anthropic deemed too dangerous to release two months ago. Guardrails keep it out of high-risk applications, but the capability jump is real — early testers describe it as a step change in AI ability, handling complex tasks that would have choked older models.

The Capacity Crunch That Killed Free Lunch

Anthropic's head of growth, Amol Avasare, posted on X that limited subscriber access is due to capacity constraints: "As enough capacity comes online, we aim to make it a standard part of those [subscription] limits again." He added, "We're sprinting as hard as we can at this," but offered no timeline. After June 22, Fable 5 will still be accessible via Extra Usage, but it won't be included in any subscription tier — Pro, Max, Team, or Enterprise.

That's a hard pivot from the original Claude subscription pitch. For now, token pricing details remain unannounced, but the pattern is clear: if you want the best frontier model, you'll meter your spending.

Why Inference Costs Are Breaking the Subscription Model

Since ChatGPT launched in November 2022, consumer AI has run on a simple $20/month all-you-can-eat buffet. Heavy users got effectively subsidized by light users, and most never hit limits. But agentic AI — longer, messier, more valuable tasks — drives inference costs through the roof. The better the model, the more people use it; the more they use it, the more hardware burns.

Anthropic isn't alone. Last month Google moved Gemini to compute-based limits. Microsoft's GitHub slashed its usage caps. OpenAI ended its usage-limit multiplier trial at the end of May. Every major player is quietly migrating from "all you can eat" to "eat what you can afford."

The 'unlimited' frontier AI era was fun while it lasted — and it lasted about three and a half years.


Source: Anthropic's new AI model is powerful, dazzling-and about to get really expensive
Domain: fastcompany.com

Read original source ->

External source stays available while the OJO article and comment thread stay local.

Comments load interactively on the live page.