Vol. 1 · Edition 033Free · No paywall

Everyone Needs a Samwise

AI news · Synthesized · Opinionated · 🌿

Jul 7Jul 12Jul 19Jul 20
1st deadline2nd deadlineTonight — finalCredits begin
Product
By Sam Taylor with Samwise

On which plan keeps Fable 5 included after tonight, what the $100 credit actually buys you in practice, and the leaked model that might be following right behind.

Three extensions later, Fable 5 finally stops being free tonight.

Source lean on this story
▲ avg

Anti-AI

00

Skeptic

02

Neutral

00

Pro (practical)

01

Pro (hyped)

01

← Anti-AI · Pro-AI →

If you've been using Claude to work through a hard email, prepare for a difficult conversation, or read through a dense document in the past few weeks — and if you've been on a paid Claude plan — there's a good chance you were talking to Fable 5, Anthropic's most powerful model, without paying anything extra for it. Tonight at 11:59 PM Pacific time, that arrangement ends.

Fable 5 launched in June — got briefly suspended under US export controls, then restored in early July with a promise: all paid subscribers would get free access while Anthropic built enough computing infrastructure to support it at full commercial scale. That access was supposed to end July 7. Then July 12. Tonight is extension number three, and Anthropic has said this is the last one. Which you'd be more inclined to believe if they hadn't said something similar the previous two times.

What Fable 5 actually is

Quick translation, because "Fable 5" and "Sonnet 5" aren't self-explanatory names.

Think of Claude Sonnet 5 as a very capable, very fast assistant who handles 90% of tasks well. Summaries, writing help, research questions, back-and-forth thinking. It's good at this. Fable 5 is the more deliberate, more careful version — better at catching what it doesn't know, better at reasoning through genuinely complicated problems, better at long documents where missing one thing in section four changes the answer entirely. The difference between a capable contractor and the senior partner who gets called in when the project gets complicated.

For most things most people do with Claude on most days, Sonnet 5 is enough. The difference shows up on hard problems. That's when it matters.

What changes tonight, by plan

Here is the clearest version I can give you.

Claude Pro ($20/month) or Team Standard: After midnight, Fable 5 is no longer inside your subscription. Using it will draw from a separate pool called usage credits. Good news: Anthropic is giving every Pro subscriber a one-time $100 credit to ease the transition — it shows up in your account automatically, nothing to claim. In plain terms, a typical 10-message conversation with Fable 5 might cost somewhere between $0.10 and $0.80 depending on how long and complex the exchange runs, which means your $100 credit stretches anywhere from a few weeks (if you run very long sessions daily) to many months (if you reach for Fable 5 occasionally). After the credit runs out, Fable 5 conversations draw from credits you'd buy separately.

Claude Max ($100/month) or Team Premium: Fable 5 stays in your subscription, permanently, at up to half your weekly message allowance. No extra charge. One thing does change: Max subscribers received a 50% bonus on Claude Code weekly usage limits during the promotional period. That bonus ends July 20. Your regular weekly limits return to normal.

Sonnet 5 is unchanged for every plan.

Fable 5 access after July 19 — by plan
PlanFable 5 after tonightExtra cost
Claude Pro ($20/mo)Usage credits only — $100 one-time credit to start$10/$50 per M tokens after credit
Claude Max ($100/mo)Included at 50% of weekly limits, permanentlyNone — but Claude Code boost ends Jul 20
Team StandardUsage credits only — $100 one-time credit to start$10/$50 per M tokens after credit
Team PremiumIncluded at 50% of weekly limits, permanentlyNone — Claude Code boost ends Jul 20

Source spread

Pros & cons

What's actually fine:

  • The $100 credit is real money and appears automatically. No action needed to claim it. For most Pro users who reach for Fable 5 a few times a week rather than daily, that credit stretches a genuinely long time.
  • Sonnet 5 is very good, and it's not going anywhere. Most of what you use Claude for — writing, quick research, daily questions — Sonnet 5 handles just as well. I mean that. Fable 5's edge shows up on hard problems, not routine ones.
  • Max subscribers effectively got a permanent upgrade included in their plan. Fable 5 at 50% of weekly limits is a meaningful ongoing benefit for the $100 price point, and it's now settled as a permanent feature rather than a temporary gift.
  • Anthropic says they plan to bring Fable 5 back to subscriptions when compute allows. That's been the consistent message. I don't have reason to disbelieve it, though I have no timeline for it.

What deserves a side-eye:

  • Three extensions. The pattern matters more than the individual deadline. Anthropic set July 7 as the final date, then moved it. Set July 12 as final, then moved it. The third "final" announcement carries less weight than it would have if this were the first one. Tonight's date might hold. It might not.
  • The credits price — $10 per million input tokens and $50 per million output tokens — is exactly double Opus 4.8's rate. Fable 5 is Anthropic's most expensive model by a meaningful margin. Regular heavy users will feel this.
  • "Will return to subscriptions when compute capacity allows" has been the message since early July. What "when capacity allows" means in practice, and how long that takes, is not stated anywhere.

The mystery in the background

On July 8, an unannounced model called "Claude Honeycomb EAP" appeared in Cursor's model selection menu and vanished within hours. The leaked specs that developers captured before it disappeared: a 1 million token context window, an "xhigh" reasoning setting, and a safety fallback that routes to Claude Opus 4.8 when needed.

That last detail is the one the community has focused on. If Honeycomb routes down to Opus 4.8 rather than Fable 5, it suggests Honeycomb sits above Fable 5 in capability. The obvious interpretation: Honeycomb EAP is an early preview of Opus 5, Anthropic's next flagship model.

What Anthropic has said about any of this: nothing. No confirmation, no denial, no comment. There is no Honeycomb or Opus 5 model ID in any public API documentation. Everything above is community reading of a brief, possibly accidental appearance.

Whether Opus 5 ships this week, this month, or this year is genuinely unknown. Plans based on the assumption would be premature.

$100
One-time credit Anthropic is giving Pro subscribers to cushion Fable 5's move to usage pricing — no action needed to claim it

→ Source: Anthropic via TechTimes

What to do about it

  • Tonight: do nothing special. Your free access runs until 11:59 PM PT. Use Claude as normal. The $100 credit will appear in your account automatically — nothing to claim or activate.
  • Tomorrow, find out which plan you're on. Log into claude.ai and check your account settings. Pro and Team Standard users have credits; Max and Team Premium users have Fable 5 included. The easiest way to tell: if Fable 5 costs extra, you're on Pro or Team Standard.
  • For everyday tasks, switch to Sonnet 5. When you start a conversation, you can select which model you're talking to. For writing, research, and daily tasks, Sonnet 5 is the smarter default — it won't touch your credits and handles most things just as well.
  • Reserve Fable 5 for genuinely hard problems. Long documents where you need to catch everything. Complicated decisions where you've already tried Sonnet 5 and the answer felt incomplete. Legal or financial questions that need careful reading. That's when the difference pays off.
  • Hold off on any Opus 5 planning. Nothing is confirmed. If and when it arrives, anthropic.com/news will be the first place it's official.

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.