Anthropic released the model they were holding back

Anthropic dropped Claude Fable 5 this morning.

Quick backstory so this makes sense. In April, Anthropic built a model called Mythos that was so capable at finding software vulnerabilities they refused to release it. They only gave it to a small group of partners like Apple, Google, JPMorgan, and AWS so those companies could use it to patch their own systems before anyone else got access.

Fable 5 is that same model, released to the public with guardrails. If you ask it something in a high-risk area like cybersecurity or biology, it quietly hands the question to Opus 4.8 instead. Anthropic says that happens in under 5% of sessions, so for business use you will likely never notice.

What you should know as an operator:

It sits above Opus. This is the new top tier. The longer and more complex the task, the bigger the impact. Multi-step projects, massive documents, and agent workflows, that is where you will actually feel the difference.
The testing window is right now. It is included on Pro, Max, Team, and Enterprise plans through June 22. After that, heavy usage may require credits until they expand capacity. If you want to pressure-test it on real workflows, the next two weeks are your window. (I assume that's why Anthropic is giving everyone double token usage right now.)
It compresses time. Early reports are showing heavy analysis and coding work that usually takes weeks finishing in days. For those of us running AI org charts, this completely changes the scope of what a single assistant can own.

I wanted to test it out myself to see if there's a difference compared to Opus. First prompt I ran this afternoon: a full outside audit of my Claude Cowork workspace across different businesses and my life areas. It spun up 4 sub-agents and ran 176 steps in parallel - completed in about 5 minutes.

The exact prompt is in the first comment. Steal it and run it on your own setup.

If you end up using Fable for something cool, share your wins below.

One more thing. This Friday, 11am MT (1pm ET), I'm doing a full deep dive on Fable 5 inside the AI Systems Lab. We'll go through strategies to maximizing your output with Claude CoWork, the workflows I'm migrating, and how to think about building an AI org chart that supports your business growth.

Comment "Fable 5" and I'll send you the details for the private session this Friday.

9 comments