Anthropic's Best AI Is Free for 12 Days

It's Wednesday, and I'm in your inbox with a 12-day countdown. Anthropic shipped its smartest model yesterday and you can use it for free until June 22.

---

Why this is a bonus edition

I don't normally do midweek sends. But this is one of those moments where the calendar matters more than the cadence: Anthropic launched Claude Fable 5 on Tuesday, and they're giving every Pro, Max, Team, and Enterprise subscriber free access to it through June 22. After that, you'll need usage credits to keep using it, at roughly double the price of Opus.

If you've been on Claude any time in the last year, this is the most consequential week to actually open the app.

---

What Claude Fable 5 actually is

Fable 5 is the safe, generally-available version of Anthropic's new top-of-the-stack model, Mythos 5. Same brain, but Fable has guardrails: about 5% of the time, when you ask it something in a sensitive area (cybersecurity, biology, chemistry), the response quietly falls back to Opus 4.8 so the most capable model doesn't end up being misused. For 99% of the work most of us do (emails, decks, analysis, research, writing), you'll never feel the handoff.

What makes it interesting isn't the safety routing. It's the benchmarks.

Fable 5 is state-of-the-art on nearly every benchmark Anthropic tested. A few that should make office professionals sit up:

• Hebbia Finance Benchmark: highest score ever recorded. Hebbia's test is built around the actual work analysts do (parsing 10-Ks, comparing comps, tracing footnotes), not toy problems.
• Scientific reasoning: when researchers were asked to compare novel hypotheses generated by Fable 5 against Opus-class hypotheses, they preferred Fable 5's about 80% of the time.
• Legal review: matches or exceeds Anthropic's previous models on long-document reasoning and citation-checking.
• Engineering case study: Stripe reportedly compressed months of engineering work into days using it on real production tasks.

The single biggest practical upgrade: long-horizon reasoning. Fable 5 holds a coherent argument across many more turns and many more pages of input than anything in the Opus line. If you've ever hit a wall around the 30th turn of a long chat where Claude started losing the thread, that wall just moved.

---

What knowledge workers should actually do with these 12 days

Here is where I think people will leave the most value on the table. The temptation is to ask Fable 5 the same questions you've been asking Opus and notice it gives slightly better answers. That's not the unlock.

The unlock is to bring it your hardest, most-deferred task. The one you've been putting off because the AI couldn't quite do it yet. Examples I'd actually try:

• The 60-page contract or lease you've been meaning to summarize, with a specific brief: "find every renewal clause, every escalation, every exit condition, and tell me what I would actually be agreeing to."
• The quarterly board update you usually spend a Sunday on: paste in the last six weekly updates and ask for a clean six-slide story with risks, asks, and a timeline.
• The competitive analysis your team keeps almost-finishing: paste in five competitor sites, three of their press releases, and ask Fable 5 to identify the strategic gap you'd be best positioned to fill.
• The data set you haven't had time to look at: drop in the CSV or screenshot and ask it to walk you through the three most interesting things in there, with the reasoning shown.

You're not testing the model. You're using the 12 days to learn whether it actually changes what you can do.

---

The June 22 cliff (and what it means for you)

Starting June 23, Fable 5 is no longer included on Pro, Max, Team, or Enterprise plans. You'll keep your subscription and your Opus 4.8 access, but using Fable 5 will require usage credits at $10 per million input tokens and $50 per million output tokens. That's double Opus 4.8.

In real money: a heavy day of Fable 5 use (long documents, multi-turn reasoning, vision-heavy tasks) could run a knowledge worker $5 to $15 in credits. A casual day, well under a dollar. For most subscribers, this is genuinely affordable as an "I'll use it when the task is worth it" tier, but it stops being the model you reach for by default.

My take: the price step is Anthropic's polite way of saying we built something that costs us a lot to run, so we need the people who use it most to pay for it. That's a fair trade. But it makes these next 12 days the only window where you can use the smartest model in the world as your default without thinking about cost.

If you're an office professional or a leader, block 90 minutes this week and put Fable 5 against your hardest task. After June 22, you'll know whether it's worth paying for, instead of guessing.

---

Pro Tip

---