What is the difference between Fable 5 and Mythos 5?

Under the hood, it is the same model. Fable 5 is the public version with safeguards: sensitive requests (biology, chemistry) fall back to Opus 4.8 — affecting fewer than 5% of sessions. Mythos 5 is the version without some of those limits, restricted to a small group of vetted cyberdefenders and infrastructure operators through Project Glasswing.

How much does Claude Fable 5 cost?

$10 per million input tokens and $50 per million output tokens via the API — double the price of Opus 4.8 ($5/$25). It is included free for Pro, Max, Team and Enterprise subscribers from June 9 to June 22, 2026; after that, usage draws on credits.

Should you switch to Fable 5?

During the free window, yes: it is the chance to test the top of the line at no extra cost. After that, it is an economic question: for most everyday tasks (emails, product sheets, simple analysis), Opus or Sonnet do the job at half the price or less. Save Fable for the hard problems where the marginal intelligence is worth double the price.

Fable, Mythos, Glasswing: why Anthropic released two versions of the same model

Q: What is Claude Fable 5?

The most powerful model Anthropic has ever made publicly available, launched June 9, 2026. It is the first model in a new “Mythos” class, above the Opus line. It is state-of-the-art on nearly every benchmark: coding, knowledge work, vision, scientific research.

1. What happened on June 9

Days after publicly warning that AI was becoming "too dangerous" to release without precautions, Anthropic launched… its most powerful model. The paradox is only apparent: it's the same decision, seen from both sides. Claude Fable 5 shipped on June 9, 2026, presented as exceeding "any model ever made generally available" by Anthropic — state-of-the-art on nearly every benchmark: software engineering, knowledge work, vision, scientific research.

The news isn't just about power: it's the lineup structure that changes, and the way Anthropic handles the risk that comes with it.

2. Mythos: a class above Opus

For years, the Claude lineup has come in three sizes: Haiku (fast, cheap), Sonnet (balanced), Opus (top of the line). I cover all of this in What is Claude. Fable 5 adds a fourth floor: the Mythos class, above Opus.

Before

3 sizes

Opus — most capable
Sonnet — the balance
Haiku — the fast one

Since June 9

4 floors

Mythos — the class above (Fable 5)
Opus · Sonnet · Haiku

Fable 5 is not an "Opus 4.9": it's the first model of a brand-new class.

On the technical side, for those who care: a one-million-token context window, up to 128,000 output tokens, and "adaptive" reasoning the model doses itself based on difficulty.

3. Why two versions of the same model

This is the heart of the story. Under the hood, Fable 5 and Mythos 5 are the same model. The difference is the safeguards:

Fable 5, the public version, ships with reinforced protections around cybersecurity and biology. Concretely, sensitive requests (biology, chemistry) are redirected to Opus 4.8, which answers instead. Anthropic says the tuning is deliberately cautious — it sometimes catches harmless requests — but triggers in fewer than 5% of sessions.
Mythos 5, the version without some of those limits, is not for sale: it's reserved for a small group of cyberdefenders and critical-infrastructure operators, vetted one by one, under Project Glasswing.

The logic: a model at this level is genuinely useful for defending systems (finding flaws before attackers do) — but the same capabilities can attack. Rather than choosing between "publish everything" and "publish nothing", Anthropic splits the difference: the public gets the power with brakes, vetted defenders get the power without them.

4. What it changes for you

Three concrete things:

The price. Fable 5 costs $10 per million input tokens, $50 per million output — double Opus 4.8 ($5/$25). For orders of magnitude, see How much does AI cost.
The free window. From June 9 to 22, 2026, Fable 5 is included at no extra cost for Pro, Max, Team and Enterprise subscribers. It's the moment to test the top of the line on your real cases — exactly what I did: Fable 5 reviewed this site the day after its release.
The right reflex. After June 22, the golden rule doesn't change: the right model is the cheapest one that does the job. Emails, product sheets, routine analysis → Sonnet or Opus are enough. Save Fable for the hard problems where the marginal intelligence is worth its price.