Anthropic tests unreleased Claude Fable 5 and Mythos 5 checkpoints

Claude Mythos will be released today (June 9th), according to leaks everywhere.

The interesting question is whether they'll also update Sonnet or Haiku.

Not that Mythos isn't enough for me, I'm just curious why the smaller models are currently getting so little attention.

Anyway, it will probably be called Claude-5-fable.

Fable 5 is state-of-the-art on nearly all tested benchmarks, with exceptional performance in software engineering, knowledge work, scientific research, and vision.

The longer and more complex the task, the larger Fable 5’s lead over our other models.

Releasing a model this capable comes with risks. Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage.

Queries on a narrow range of topics will instead receive a response from our next-most-capable model, Opus 4.8.

Claude Fable 5 (Mythos) is finally out!

This is what I was looking for!!!!!!!!!!!!!!!!!!!!

Claude Fable 5 is out

Claude Fable 5 is available everywhere today. Claude Mythos 5 is restricted to Glasswing partners until we expand our trusted access program.

https://www.anthropic.com/news/claude-fable-5-mythos-5

I'm so excited to sell some feet pics to the highest bidder

For a small group of cyber defenders and critical infrastructure providers, we are also launching Claude Mythos 5.

Mythos 5 shares the same underlying model as Fable 5, but with the safeguards lifted in some areas.

Fable 5’s safeguards detect requests related to cybersecurity, biology and chemistry, and distillation. Users are informed whenever a fallback occurs—on average in less than 5% of sessions.

We’ll keep refining the safeguards to reduce false positives.

"dad, what did mythos cost?"

Soon, we intend to expand access to Mythos 5 through a broader trusted access program, both for defensive cybersecurity work and biomedical research.

@claudeai feels like

"dad, how much did mythos actually cost?"

@claudeai its cooking time boysssssssss

Apparently tomorrow 🤷‍♂️

7 hours ago

claude-fable-5 👀

@scaling01 he knew

@claudeai Very excited for folks to try this one out!

@claudeai A BEAST of a model.

here's our vibe check from a week of testing:

Genuinely impressive, but most small businesses won't touch a fraction of the use cases it’s actually built for. When it comes to building business automations like answering the same ten questions all day, creating generic invoices, Opus 4.8 is already overkill. The win is matching the model to the job, not running the most powerful one on work a cheaper model handles fine.