Anthropic is set to release a new model called Claude Fable, a safeguarded public version of the internally code-named "Mythos," on June 9, 2026 (US time), with pricing reportedly set at roughly twice that of Claude Opus.
June 9, 2026 · Anthropic
Anthropic to Release "Claude Fable" — a Tamed Slice of Its Frontier "Mythos"
A sanitized, safety-hardened version of the withheld Mythos model is expected to launch publicly at roughly twice the price of today's top-tier Opus — a steep cut from the invite-only preview's pricing.
~2×
Fable's expected price vs. Claude Opus
244+
Pages in the Mythos system card
1M
Token context window supported
Three Tiers, One Family — Priced per 1M Tokens (output)
Output-token price drawn to scale. Mythos Preview stays invite-only.
$25
Claude Opus
~$50
Fable (public)
$125
Mythos Preview
Frontier Benchmarks — Mythos Preview
Why the full model stays locked away
Mythos can autonomously discover and exploit unpatched zero-day vulnerabilities — judged too risky for general release.
Zero-days in the thousands
External AISI evaluations confirmed strong multi-stage attack and CTF performance.
Glasswing only
Preview shared with Amazon, Apple, Google, Microsoft, Nvidia, CrowdStrike.
Fable = guardrailed
Hardened against cyber and bioweapon misuse, tuned toward defense.
Developer optimism
Top-tier reasoning, coding and agent skills — "clearly above Opus" and seen as transformative for long-horizon tasks.
Lingering caution
Higher price may shut out individuals; real-world utility hinges on guardrails — and on how much power the restricted version keeps.
No official announcement yet — pricing and specifications may change.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…
