Skip to content

AI

Anthropic's Fable 5.1 drops the leash, keeps the watchers

Fable 5.1 is live today for anyone with an API key, while Mythos 5.1 stays locked behind a partner badge. The real twist is who gets to watch for misuse: now it's your call.

TL;DR

  1. Fable 5.1, the unrestricted model, is live today on cloud platforms and the Anthropic API.
  2. Mythos 5.1 stays limited to registered partners in cybersecurity or life sciences research.
  3. Zero data retention arrives, and Enterprise Frontier Safeguards lets clients run misuse monitoring themselves this fall.

What happened

Anthropic released Fable 5.1, the unrestricted version of its model. It is available starting today through cloud platforms and the Anthropic API. No waitlist drama, no partner handshake.

Its sibling, Mythos 5.1, is a different story. As with the previous Mythos, access is reserved for registered Anthropic partners doing cybersecurity or life sciences research. The velvet rope is still up.

The bigger structural change is privacy. Anthropic now supports zero data retention, meaning clients can run its models on their own infrastructure with no data flowing back out. Fable was previously off the table for this because of security concerns.

To fix that, Anthropic is introducing a high-privacy service called Enterprise Frontier Safeguards, rolling out to users in the fall. The system still watches for misuse by agents or humans. The difference is that clients decide how that monitoring works.

Anthropic also told customers their data had not been inappropriately accessed, and repeated that it does not train on business customer data unless the customer explicitly agrees, and does not plan to start.

As usual for an Anthropic launch, the company reports record results on benchmarks including Terminal-Bench 4.0 for command-line coding and Humanity's Last Exam for general reasoning. Those are vendor numbers, so treat them as a pitch until someone else runs the tests.

Why it matters

The headline is access. An unrestricted frontier model that you can call from your usual cloud console is a very different product from one gated by research credentials. Anthropic is splitting its lineup cleanly: broad and open on one side, specialist and supervised on the other.

The quieter story is who holds the monitoring dial. Enterprises that could not touch Fable because their data could not leave the building now have a path. Misuse watching does not disappear, it just moves into the customer's hands.

That is a real shift in how a frontier lab handles trust. The model still has a chaperone. You simply pick the chaperone's hours.

One caveat worth keeping in mind: Enterprise Frontier Safeguards is a fall rollout, not a today feature. The zero data retention pitch is a promise with a calendar attached.

What to do with it

  • Try Fable 5.1 today.It is on cloud platforms and the API, so point an existing workload at it and compare the output against your current model.
  • Check your data rules.If privacy blocked you from Fable before, flag Enterprise Frontier Safeguards for your security team ahead of the fall rollout.
  • Verify the benchmarks yourself.Records on Terminal-Bench 4.0 and Humanity's Last Exam are Anthropic's claims, so run your own tasks before you rewrite a roadmap.
  • Apply for Mythos only if it fits.It is for registered partners in cybersecurity or life sciences research, so skip it if that is not you.

Keep up, sapiens. The models got a privacy upgrade before most of us did.