Anthropic Fable 5.1 AI model is cheaper and better at coding
Anthropic released Claude Fable 5.1 on Tuesday, a new version of its flagship AI model that reduces operating costs and improves coding performance. The company also released Claude Mythos 5.1, an identical model with different safeguard levels available only to vetted partners in cybersecurity and life sciences.
The most significant cost change is a 75% reduction in cache read pricing — the fee charged when the model reuses context it has already processed — dropping to $0.25 per million tokens. For typical workloads, Anthropic said the change will reduce overall costs by around 25% compared to Fable 5. For complex coding and highly agentic tasks, savings could reach approximately 45%. The base price of $10 per million input tokens and $50 per million output tokens remains unchanged.
Dianne Penn, who leads product management for research and labs at Anthropic, told Bloomberg that cache reads can make up a majority of tokens consumed during the most demanding workloads, making the pricing cut especially significant for enterprise customers.
Alongside the pricing changes, Anthropic updated its safety classifiers to reduce false positives. The company said Claude Code users can expect roughly 60% fewer cybersecurity-related interventions per session compared to the previous safeguards on Fable 5. Fable 5.1 can now be used to identify software vulnerabilities, though not to develop exploits. In biology, the company said its safeguards now fire 85% less often for benign requests related to elementary biology and medical questions relative to those that launched with Fable 5.
Anthropic also introduced Enterprise Frontier Safeguards, a new system that allows business customers to store data on their own cloud infrastructure rather than on Anthropic's servers, while still enabling the company to monitor for misuse. The feature is rolling out in phases beginning later this fall. Until it is available, eligible customers can use Fable 5.1 with zero data retention, the company said.
Fable 5.1 also showed performance improvements across several benchmarks, including a score of 55.8% on Terminal-Bench 4.0 for agentic coding, up from Fable 5's 42.0%, and 52.6% on Terminal-Bench-Science 0.1, compared to Fable 5's 24.7%.
Fable 5 and Mythos 5 launched in June as the most powerful models in Anthropic's lineup; access to Mythos was kept tightly limited from the start given concerns about its cyber and biological capabilities. The models were later subject to a brief government-ordered shutdown before being reinstated in July after Anthropic developed a new safety classifier to address a vulnerability documented by Amazon researchers.
Claude Fable 5.1 is available on the Claude API and cloud platforms including Amazon Web Services, Google Cloud, and Microsoft Azure.
