Anthropic releases Fable 5.1 and tightens Mythos access
The launch paired cheaper pricing with invisible text watermarks and a private-preview detection tool as Anthropic widened the gap between its public and restricted models.
On 1 September, Anthropic released Claude Fable 5.1 for general access and kept the more permissive Claude Mythos 5.1 configuration limited to vetted cybersecurity and life sciences users. The company said the two are the same model, but with different levels of safeguards. As covered in August, Anthropic had already started adding watermarks to newer Claude text to meet EU transparency rules.
According to Anthropic’s system card, Fable 5.1 is open to general users with extra safeguards that block certain high-risk, dual-use tasks. Mythos 5.1 uses more permissive safeguards in those areas, but direct access is limited to vetted individuals and organisations through trusted access programmes.
Anthropic said the main security change lets Fable 5.1 identify software vulnerabilities for defensive use, while exploit generation remains restricted. It said the updated cybersecurity safeguards cut false positives by 60%.
Pricing also changed. Cache-read charges fell by 75% to $0.25 per million tokens, while input and output prices stayed at $10 and $50 per million tokens. Anthropic estimated the shift would make typical workloads about 25% cheaper than Fable 5, rising to around 45% for complex coding and highly agentic tasks.
The release also sat inside the European Union’s AI Act transparency regime. The European Commission says Article 50 obligations took effect on 2 August, and although the code of practice is voluntary, the transparency requirements are legal obligations. Anthropic said models launched after that date must add a numerical watermark to help assess whether Claude likely helped produce the text.
Anthropic said a detection API is entering private preview for eligible regulators, law enforcement, media organisations, fact-checkers, researchers, educational organisations and EU civil society groups.
The company also said Enterprise Frontier Safeguards will store customer data in cloud infrastructure controlled by the customer rather than Anthropic, with a phased rollout for enterprise customers beginning later this fall.