Anthropic has announced the release of Claude Fable 5.1 and Claude Mythos 5.1, describing them as the most advanced models for coding and knowledge work. The standard input and output token prices remain at $10 and $50 per million tokens respectively. Typical workloads are expected to cost about 25 percent less because of a major reduction in cache read pricing.
Performance Benchmarks and Model Specifications
The new models show substantial gains over their predecessors across several benchmarks. Fable 5.1 scored 52.6 percent on Terminal-Bench-Science 0.1 compared with 24.7 percent for Fable 5, and it reached 31.4 percent on AutomationBench versus 17.1 percent. On Terminal-Bench 4.0 it scored 55.8 percent, while Mythos 5.1 reached 60.9 percent, and the models now have a knowledge cutoff of June 2026 with an unchanged one million token context window.
Safeguards and Model Versions
Fable 5.1 and Mythos 5.1 are the same underlying model but differ in their safety filters. Fable 5.1 is generally available for paid Claude users and API customers, while Mythos 5.1 is currently restricted to a set of vetted United States organizations. The cybersecurity filter now permits vulnerability discovery but continues to redirect exploit generation and penetration testing to the Opus models.
Pricing and Cost Savings
Anthropic has kept the base price unchanged at $10 per million input tokens and $50 per million output tokens. However, the cost of cache reads has fallen from $1.00 to $0.25 per million cached tokens. Based on four weeks of usage data, Anthropic estimates this reduces typical workloads by around 25 percent and highly automated tasks by up to 45 percent.
Enterprise Privacy and Regulatory Compliance
Later this year Anthropic will introduce Enterprise Frontier Safeguards so companies can host data in customer-controlled cloud infrastructure while still allowing misuse monitoring. Until then eligible customers can operate Fable 5.1 with a zero data retention policy. The model also includes an invisible statistical watermark to comply with the EU AI Act, with a detection tool in private preview for regulators and researchers.
Anti-Distillation Measures and Early Reports
Anthropic has adjusted API behavior to make model distillation harder. New API accounts can no longer edit prior messages in a multi-turn conversation while preserving the model's underlying reasoning data. The company also shared testimonials from early adopters, including a report that Fable 5.1 identified a rare crash caused by a vendor library bug, although these examples were curated by Anthropic.
Scientific Research Applications
Anthropic highlighted early scientific results as evidence of rising research capabilities. Mythos 5.1 designed protein binders with a hit rate of nearly 50 percent across twelve targets, far above the typical 10 to 15 percent range. The company also cited a high-resolution Venus elevation map and GPU optimizations that sped up open source biology models by up to 2.5 times.
Safety Testing and Filter Precision
Before release, two external organizations and Gray Swan tested the updated safeguards and found no critical severity jailbreak. Anthropic says the cybersecurity filter now triggers about 60 percent less often per session in Claude Code. Biology safeguards fire 85 percent less often on ordinary medical questions, while life sciences research requests are still directed to Opus models.
Claude Fable 5.1 is available immediately on major cloud platforms and through the Claude API, while access to Mythos 5.1 remains limited to trusted programs. The update combines higher benchmark performance with lower effective costs and more precise safeguards for enterprise and scientific users. Anthropic is positioning the release as both a practical improvement for everyday workloads and an early step toward broader AI contributions to scientific discovery.