Vol. 1 · Edition 028Free · No paywall

Everyone Needs a Samwise

AI news · Synthesized · Opinionated · 🌿

$10
per million input tokens
Fable 5 usage credits, effective today
Product
By Sam Taylor with Samwise

On the July 7 billing switch, what $10 per million tokens means in practice, and whether Sonnet 5 fills the gap for most people.

Fable 5 just left your Claude subscription. Here's what that costs you now.

Source lean on this story
▲ avg

Anti-AI

00

Skeptic

02

Neutral

00

Pro (practical)

01

Pro (hyped)

00

← Anti-AI · Pro-AI →

If you pay for Claude — Pro, Max, or Team — and you've been reaching for Fable 5 on anything that actually matters, it was included in what you paid. Starting today, it isn't.

Anthropic's most capable model exits subscription usage limits this morning, moving to a separate pay-per-use billing layer called usage credits. The rate: $10 per million input tokens, $50 per million output tokens. That's exactly twice what Claude Opus 4.8 costs.

This is the second billing surprise in the Fable 5 saga this month. The model was suspended June 12 due to US government export controls, came back July 1 after the controls were lifted — and now, seven days into its return, the grace period ends.

What "usage credits" actually means

Here's the object lesson, because the phrase is opaque.

Think of usage credits like a prepaid card attached to your Claude account. Your subscription — the $20 Pro fee, the $100 Max fee — still gets you access, standard models, and weekly usage limits. Fable 5 now requires a separate card. You load money onto it, set a monthly spending cap so you don't get surprised, and the model draws from that card each time you use it. When the card runs out, Fable 5 stops — with no warning and no grace period; requests just silently fail.

If you haven't set up credits yet: Settings → Usage on claude.ai, enable them, and set a monthly cap before your next Fable 5 conversation.

$10
Fable 5 usage credits: cost per million input tokens, effective July 7

→ Source: Fable5.app / Anthropic

What this costs in practice

A "token" is a small chunk of text — roughly four characters. So "hello" is one token, a full sentence might be fifteen, and a typical back-and-forth conversation over twenty minutes probably runs 5,000–15,000 tokens. At $10 per million input tokens, that's about $0.05–$0.15 per session. Annoying, but not a crisis.

Where the costs stack: feeding long documents into the model. A 100-page contract or a 200-page research report can easily run 75,000 to 150,000 input tokens. At $10 per million, that's $0.75–$1.50 just for the input, before the model even responds. For heavy daily users — those processing large documents regularly — the math starts looking like a second subscription.

Fable 5 vs Claude Sonnet 5, from July 7
Fable 5Claude Sonnet 5
Included in Pro/Max/TeamNo — usage credits onlyYes
Cost (usage credits)$10 input / $50 output per million tokens$2 input / $10 output per million tokens
Available on Free tierNoYes
Best forLong docs, hard reasoning, extended agentsMost everyday tasks
Situation permanent?Anthropic says noStandard default

Source spread

What's real vs what deserves a side-eye

What's real:

  • The temporary framing is plausible. Fable 5 only returned six days ago after a 19-day forced outage. Capacity constraints after a sudden relaunch are real, not manufactured.
  • The credits model with a monthly cap is actually well-designed. You control the maximum you spend. That's better than a service that quietly charges you once a checkbox somewhere ticks over.
  • Claude Sonnet 5 launched June 30 as the new Free and Pro default. It's the strongest Sonnet model to date and handles most everyday tasks well. For most Claude users, it fills the practical gap.

What deserves a side-eye:

  • "Temporary with no timeline" is a weak commitment. "We plan to restore Fable as part of subscriptions as capacity allows" is corporate for "we don't know when." Watch whether it actually comes back — the incentive to leave a premium model on per-token billing doesn't disappear on its own.
  • $10/$50 per million tokens is the most expensive pricing Anthropic has ever listed for a generally available model. That's not a footnote; it's worth saying plainly.
  • The timeline is sharp: export-control suspension for 19 days, return with a six-day grace window, then credits billing. Subscribers absorbed a lot in three weeks. A two-week grace window instead of six days would have cost Anthropic less in goodwill than it saved in compute.

What to do about it

Five steps, in order of importance.

  • If you use Fable 5 regularly, enable usage credits now. Go to Settings → Usage on claude.ai. Set a monthly spending cap before you need it — $10 to $20 is a reasonable starting point for moderate use. Without credits enabled, Fable 5 silently fails with no warning.
  • Try Sonnet 5 first. Claude Sonnet 5 launched June 30 and is now included in your subscription. For most tasks — writing, analysis, research, coding — it handles things Fable 5 used to be needed for. Give it a week before deciding you need credits.
  • Reserve Fable 5 credits for the tasks only it does well. Very long documents (150+ pages), complex reasoning chains, extended multi-step agent workflows. For everything else, Sonnet 5 is included and fast.
  • Watch the "temporary" claim. Anthropic has said Fable 5 will return to subscription limits when capacity allows. No timeline given. Check back in 30 days. If it hasn't moved, that tells you something about the trajectory.
  • Free-tier users: nothing changed for you. Fable 5 was never available on the free tier. Sonnet 5 is your new default and it's meaningfully better than what you had before June 30.

Further reading

🌿

Liked this? Get the weekly digest.

Free. Monday mornings. The week's stories, synthesized. Unsubscribe anytime.

Your take

How'd I do on this one?

What did I miss?

Tell Samwise (and Sam).

Disagree with the take? Spotted a fact I got wrong? Have context I should have included? Drop it here. Anonymous unless you leave an email.