spoonai
TOPAnthropicFable5Mythos5

Anthropic Fable 5 Switches to Usage-Credit Billing Today (June 23) — $10 Input / $50 Output per 1M

Fable 5, launched June 9 and bundled free into Pro, Max, Team, and Enterprise, leaves those plans today and moves to usage-credit billing: $10 per 1M input tokens, $50 per 1M output — exactly 2x Opus 4.8. And it's tangled up with US export controls.

·7분 소요
공유
AI 데이터센터 GPU 서버랙
Unsplash

What was free yesterday now costs money starting today

The billing for Anthropic's newest public model, Fable 5, changed today (June 23). When it launched on June 9, Fable 5 was bundled at no extra cost into Pro, Max, Team, and seat-based Enterprise plans. That free-inclusion window ended on June 22, and as of today it moves to usage-credit billing — meaning what was inside your subscription yesterday now bills separately per use.

The pricing isn't trivial: $10 per 1M input tokens, $50 per 1M output tokens. That's exactly double Anthropic's flagship Opus 4.8 ($5/$25). And this is more than a price reshuffle — the model is entangled with US government export controls, so "billing change" only scratches the surface. Let's walk through it.

Who and what — Fable 5 and Mythos 5

Mythos 5 is Anthropic's most powerful frontier model. Fable 5 is the version of Mythos 5 made publicly accessible. If Mythos is the internal, top-tier engine, Fable is the face that ships it through the API and subscription plans. At launch on June 9, TechCrunch described Fable 5 as "the version of Mythos the public can access today."

The name may ring a bell — Fable 5/Mythos 5 dominated June here: launch (6/9), worldwide access shutdown (6/12), export-control talks (6/17-19), and the trial window closing (6/22). Today's piece is the final puzzle piece in that run: the billing-model switch.

Chris Ciauri (Anthropic's MD for international) matters here. He sits where Fable 5's global policy and access issues get coordinated — the key person sorting out "who, where, and at what price" can use the model when export controls and pricing collide at once.

What changed — the price list and timeline

The core is the shift from "bundled free" to "pay per use." The price table makes it concrete.

Item Detail
Input $10 per 1M tokens
Output $50 per 1M tokens
Cached read $1 per 1M tokens
Batch $5 / $25 (50% off standard)
Benchmark Exactly 2x Opus 4.8 ($5/$25)

The timeline ran like this:

Date Event
2026-06-09 Fable 5 launches, bundled free into Pro/Max/Team/Enterprise
2026-06-12 US export controls take Fable 5/Mythos 5 offline worldwide
2026-06-22 Free-inclusion window ends
2026-06-23 Usage-credit billing begins (today)

Two takeaways. First, Fable 5 — free inside subscriptions until yesterday — drops out today, and using it requires buying credits. Second, at 2x Opus pricing, it's positioned not as a "use it freely" model but a "pick it when you need it" premium one. Batch runs at half price, so cost-sensitive bulk work belongs in batch now.

Who wins — why Anthropic did this

Anthropic gains cost recovery and demand control. Fable 5/Mythos 5 are compute-heavy top-tier models; bundling them unlimited inside subscriptions means bleeding money. Pull users in with a free window, then convert to billing to keep only the demand that truly values it — textbook premium SaaS. Pricing at 2x Opus also signals "the top model is worth the premium."

Enterprise and developer users split. Shops that sampled it free and now want Fable 5 only on workloads that need it may find pay-per-use more rational. But individuals and small teams who used it unlimited on a flat subscription face a sudden cost. With output at $50, jobs that generate long answers rack up fast.

On the competitive side, this price invites a value comparison against OpenAI's and Google's top models. Users now compute whether Fable 5 is worth 2x. The comparison nobody made while it was free starts the moment billing kicks in.

Past parallels — wins and losses

"Free, then billed" is a familiar IT playbook. Cloud and SaaS firms routinely give new features away to lock users in, then monetize. Done well, users already embedded in the workflow convert to payment. OpenAI has released new models free for a window and shifted to API billing many times, generally keeping its core users.

But failures exist too. Users acclimated to free push back and churn, or a rival slips in claiming "we're cheaper." At 2x Opus, expect resistance — "do I really need to pay double for Fable 5?" Whether the free window etched in enough irreplaceable value will decide the outcome.

And unlike a normal monetization, this one carries an export-control variable. A June 12 US directive blocked foreign-national access and briefly took the model offline worldwide. If billing starts but "who can use it" stays restricted by region and nationality, the business model itself wobbles — a risk more fundamental than price.

Competitor counter-play

The most direct rival is OpenAI. As Anthropic bills its top model at 2x Opus, OpenAI can counter with "similar performance cheaper" or "more for the same price." The moment a free window ends and users open their wallets is exactly when a rival pitches "switch to us."

Google Gemini has a card too. Gemini's recent "Flash" line pushes "fast, cheap, frontier-class," so if Fable 5 starts billing high, "Gemini is enough" cost logic can land — sharper the more cost-sensitive and bulk the workload.

Open-weight players are a variable as well. If Chinese open models like DeepSeek and Qwen handle similar work for almost nothing, the split between "needs a frontier top model" and "an open model suffices" accelerates. Fable 5's monetization sharpens that fork.

So what actually changes

If you used Fable 5, re-check your cost structure today. With the bundle gone, pick which jobs truly need Fable 5 and spread the rest to Opus or cheaper models. Route bulk work through 50%-cheaper batch.

If you're an enterprise decision-maker, multiply that "$50 output" by your workload. Services with heavy long-document generation can balloon faster than expected. Also confirm whether the export-control region/nationality limits affect your org.

If you watch the AI market, this signals that frontier top-tier models are now getting properly priced. The launch-free → lock-in → monetize cycle is speeding up, and users increasingly compute whether a job really needs the top model. Choosing a model is becoming a cost strategy.

🥄 Three Things You're Probably Wondering

— So can I not use Fable 5 anymore? You still can — it just left the subscription plans. Buy usage credits and it's available. But at $10 input / $50 output, it's no longer a "use it freely" model. Pick it only for the jobs that need it.

— Am I blocked by export controls? Too early to say — it depends on region and nationality. Access was briefly blocked worldwide after the June 12 directive and is still settling, so check whether it's actually live for your location.

— Is it worth paying 2x Opus? Depends on the job. For hard tasks where top-tier performance clearly lifts quality, it can pay off; for ordinary work, Opus or a cheaper model is often enough. It's exactly the point where people who compared during the free window disagree.

Sources

Pricing and policy are as of announcement and may change.

관련 기사

무료 뉴스레터

AI 트렌드를 앞서가세요

매일 아침, 엄선된 AI 뉴스를 받아보세요. 스팸 없음. 언제든 구독 취소.

매일 30개+ 소스 분석 · 한국어/영어 이중 언어광고 없음 · 1-클릭 해지