TL;DR
Anthropic released Claude Fable 5.1 and Mythos 5.1 with reduced cache costs and improved coding and scientific research capabilities.
Anthropic released Claude Fable 5.1 and Mythos 5.1, its latest flagship models for coding and knowledge work, while slashing cached token pricing by 75%. The update targets enterprises running AI agents on complex, multi-step tasks involving sensitive data.
Fable 5.1 is generally available with built-in safeguards, while Mythos 5.1 is restricted to approved cybersecurity and life-sciences organizations via a trusted-access program. Both models share the same underlying architecture but differ in safety restrictions.
The company claims Fable 5.1 delivers stronger performance on software projects, code reviews, and multistep scientific workflows. It also helps design experiments and interpret detailed tables and diagrams.
On internal benchmarks, Fable 5.1 scored 55.8% on Terminal-Bench 4.0, up from 42% for Fable 5. Mythos 5.1 reached 60.9% under relaxed safety settings. On Terminal-Bench-Science 0.1, Fable 5.1 achieved 52.6%, more than double Fable 5's 24.7%.
Pricing changes include a 25% reduction for typical workloads and up to 45% for heavily agentic tasks. Cache reads now cost $0.25 per million tokens, down from $1. Base input and output rates remain at $10 and $50 per million tokens respectively.
Anthropic also introduced Enterprise Frontier Safeguards (EFS), letting customers store data on their own infrastructure. EFS rolls out this fall across Claude Code, Claude Enterprise, and major cloud platforms.
In scientific testing, Mythos 5.1 reportedly designed high-affinity protein binders with a 50% hit rate across 12 targets. Fable 5.1 helped build a new elevation map covering a third of Venus's surface using NASA Magellan data.
These benchmark results are Anthropic's own and should not be treated as independent proof of superiority.
The market reaction has been cautious, with developers weighing cost savings against safety trade-offs. Early adopters are testing Mythos 5.1 in controlled environments.
For practitioners, the real test lies in how these models perform under extended agentic workloads. Cost efficiency gains may drive adoption, but trust in safety controls remains key.
What happens when open models match or exceed proprietary performance at lower cost?
FAQ
Q: What is the difference between Fable 5.1 and Mythos 5.1?
A: They share the same architecture. Fable 5.1 has standard safeguards and is publicly available. Mythos 5.1 has fewer restrictions but is limited to trusted-access programs.
Q: How much did cache pricing drop?
A: Cache read costs fell 75%, from $1 to $0.25 per million tokens.
Q: Are the benchmark scores independently verified?
A: No. All scores cited by Anthropic are self-reported and not independently validated.
Q: When will Enterprise Frontier Safeguards launch?
A: EFS begins rolling out this fall across Claude Code, Claude Enterprise, and AWS, Google Cloud, and Microsoft Azure.
Tags: artificial intelligence, machine learning, AI models, coding, scientific research
Risk Level: LOW
Legal Notes: []
About the Author
Guilherme A.
Former dentist (MD) from Brazil, 41 years old, husband, and AI enthusiast. In 2020, he transitioned from a decade-long career in dentistry to pursue his passion for technology, entrepreneurship, and helping others grow.
Connect on LinkedIn