Anthropic has released Fable 5.1, an unrestricted version of its most advanced model, alongside a twinned release called Mythos 5.1 reserved for registered partners in cybersecurity and life sciences. Fable 5.1 is live now on major cloud platforms and through the Anthropic API, with a lower token cost and fewer false-positive restrictions from its safeguards compared with earlier versions.
What changed in Fable 5.1
Two practical shifts define this release. First, token pricing is lower, which changes the math for any team running the model at scale. Second, the safeguards that previously blocked legitimate requests have been tightened to cut down on false positives, so fewer useful prompts get refused by mistake.
For enterprises, the release comes with a privacy posture that addresses a long-standing friction point. Anthropic states that it has never trained on enterprise data without explicit permission, a baseline that matters most to regulated buyers.
How the two versions differ
Fable 5.1 is the general-purpose build. It is available now through cloud platforms and the Anthropic API. Mythos 5.1 is the partner-only build, restricted to registered organizations working in cybersecurity or life sciences.
The split mirrors how Anthropic has handled its highest-capability models before, with a more controlled version for sensitive use cases and a wider one for everything else. Both versions are built on the same underlying capability, with Mythos tuned for workflows where access controls are tighter.
Enterprise privacy and where models can run
Zero data retention is available for Fable 5.1: clients can run the models on their own infrastructure, and nothing flows back out. That removes a recurring objection from security teams who want to use frontier models inside private environments.
A new high-privacy service called Enterprise Frontier Safeguards arrives in the fall. The service was previously unavailable for Fable on security grounds, so its arrival opens a category of workloads that could not use the model before. Anthropic will continue misuse monitoring, but clients control how monitoring happens inside their own deployments.
What the models produced before release
Three novel scientific findings came out of the models during pre-release testing. Among them:
- A GPU optimization, the kind of internal-performance win that can lower compute costs across a fleet.
- A high-resolution map of Venus built from existing photographs, a useful test of how the model handles scientific reconstruction tasks from legacy data.
Findings like these show what the models can contribute before any commercial deployment, and they give buyers concrete examples to evaluate.
What the system card reports
Anthropic published a system card with the release, and the safety picture is mixed. Mythos is rated low-risk for automated AI development and shows a slight regression on misaligned behavior when compared against Opus 5. Both points are worth flagging for any team using the model inside a research or automation pipeline.
The card also lists benchmark coverage including Terminal-Bench 4.0 and Humanity’s Last Exam, standard tests used to track model progress across coding and reasoning. Anthropic has not published headline scores from those runs in this release.
Why the privacy unlock matters for buyers
Frontier-model buyers typically split into two camps: teams that want raw capability and can route requests through an API, and teams that need the model inside a private environment because of data sensitivity. Fable 5.1 now serves both, with cheaper tokens, fewer false-positive refusals, and a deployment path that keeps data inside the customer’s perimeter.
For teams that have been holding back on a frontier model because of data-handling concerns, the combination of zero data retention, the new Enterprise Frontier Safeguards, and self-hosted deployment makes Fable 5.1 a more practical option than earlier versions.
FAQ
What is Fable 5.1?
Fable 5.1 is Anthropic’s unrestricted top-tier model, live on major cloud platforms and through the Anthropic API, with lower token cost and reduced false-positive restrictions.
What is Mythos 5.1?
Mythos 5.1 is a partner-only version of the same model, limited to registered organizations working in cybersecurity or life sciences.
Can Fable 5.1 run inside a private environment?
Yes. Clients can run Fable 5.1 on their own infrastructure with zero data retention, and Anthropic’s Enterprise Frontier Safeguards high-privacy service arrives in the fall.
Related coverage
See our editorial disclaimer for how our articles are produced.
Run a free scan to see your AI Visibility Score, SEO rating, and local citation accuracy.