Anthropic Releases Fable 5.1 With Lower Costs and Fewer False Restrictions

Anthropic has released Fable 5.1 and Mythos 5.1, new versions of its advanced artificial intelligence models that introduce lower token costs, updated safety controls, and improved privacy options for enterprise users.

Fable 5.1 is the unrestricted version of the release and is available through cloud platforms and the Anthropic API. Mythos 5.1, the more restricted version, will only be available to registered Anthropic partners working in cybersecurity and life sciences research.

The company said the latest release includes changes designed to reduce false-positive restrictions from model safeguards while maintaining protections against misuse. The update also focuses on reducing the cost of using the models by lowering token expenses.

Anthropic Expands Privacy Controls With Enterprise Frontier Safeguards

One of the major changes in the release is Anthropic’s expansion of zero data retention capabilities, allowing customers to run Anthropic models on their own infrastructure without data leaving their environments. The development follows wider industry discussions around zero-retention AI systems and privacy-focused enterprise deployments.

Anthropic said its high-privacy service, Enterprise Frontier Safeguards, will roll out to users in the fall. The system will continue to monitor for misuse by AI agents or human users, while giving customers more control over how monitoring is carried out.

The company also assured customers that enterprise data has not been used to train its models without permission. Anthropic said it has never trained on enterprise data without explicit authorisation and will not do so in the future.

Fable 5.1 and Mythos 5.1 Improve Benchmark Performance

Anthropic said the new models achieved strong results across several evaluations, including Terminal-Bench 4.0, which measures command-line coding performance, and Humanity’s Last Exam, a benchmark designed to test advanced reasoning abilities.

The company also announced three scientific discoveries generated by the models before release. These included a custom GPU optimisation and a high-resolution map of Venus created from existing images.

As with previous Anthropic releases, the company published a detailed system card for Fable 5.1 and Mythos 5.1 explaining the models’ capabilities, limitations, and safety evaluations.

Anthropic’s Safety Review Highlights Model Behaviour Changes

The system card classified Mythos 5.1 as “low-risk” regarding concerns related to automated AI development, in which AI systems could assist in improving future AI systems. Anthropic said the model’s ability to accelerate internal AI research and development aligns with current trends.

The evaluation also found that Mythos 5.1 showed some differences compared with earlier models. Anthropic said the model was slightly more likely than Opus 5 to exhibit certain forms of misaligned behaviour, while improving in areas such as following explicit constraints and avoiding false claims about completed tasks.

According to Anthropic’s assessment, Mythos 5.1 is less likely than previous models to ignore restrictions, hallucinate inputs, or falsely claim to have completed actions. The company said the updated model still requires continued evaluation as capabilities improve.