Anthropic Unveils Claude Fable 5.1 and Mythos 5.1, Detailing Their Features

Anthropic has introduced two upgraded Claude models aimed at software development, scientific research and other knowledge‑intensive tasks.

Claude Fable 5.1 is immediately accessible to anyone using Anthropic’s platform, while Claude Mythos 5.1 is restricted to organisations that pass Anthropic’s verification programmes.

Both versions share the same core architecture, but Anthropic has applied distinct safety configurations to meet different risk tolerances.

New Model Variants and Availability

Claude Fable 5.1 is marketed as the general‑purpose offering, allowing the broad public to experiment with advanced coding and reasoning capabilities.

Claude Mythos 5.1 is positioned for approved customers in high‑sensitivity sectors such as cyber security and life sciences, where deeper model access may be required.

The safety layer on Mythos 5.1 is deliberately less restrictive, reflecting the trust placed in verified organisations.

Access to Mythos 5.1 will be granted through Anthropic’s Cyber Verification Program and Life Sciences Verification Program, each designed to vet applicants’ security and ethical standards.

Anthropic’s blog post notes that user feedback on pricing, data handling and safety shaped the design of both models.

Performance Gains and Benchmarks

Internal testing shows Claude Fable 5.1 achieving a 52.6 percent score on the Terminal‑Bench‑Science suite, more than double the 24.7 percent recorded by its predecessor, Claude Fable 5.

For comparison, Opus 5 obtained 29 percent on the same test, underscoring the leap in scientific reasoning for Fable 5.1.

In the CursorBench evaluation, Fable 5.1 reached 73.4 percent, edging out Claude Fable 5’s 70.5 percent and Opus 5’s 70 percent.

These benchmark improvements are complemented by real‑world trials conducted with investment firm Millennium.

During the Millennium experiment, Fable 5.1 correctly pinpointed a third‑party library bug as the cause of a persistent software crash.

Anthropic reports that the model has also been exercised in computational analysis, scientific research and computational biology projects.

Claude Mythos 5.1 has demonstrated a near‑50 percent success rate in designing high‑affinity proteins, a metric relevant to pharmaceutical research.

The tweet from Anthropic’s official account reads: “We’re introducing Claude Fable 5.1 and Claude Mythos 5.1. They're the world’s most advanced models for coding and knowledge work.”

Pricing, Access Programs and Use Cases

Anthropic claims Claude Fable 5.1 is roughly 25 percent cheaper than Claude Fable 5, attributing the reduction to fewer cache reads.

For workloads with heavy agentic usage, the cost advantage can rise to about 45 percent.

The lower price point is intended to make high‑performance AI more affordable for a wider range of developers and researchers.

Organizations approved for Mythos 5.1 gain the ability to run more permissive queries, which can be valuable in cyber threat analysis and advanced drug discovery.

Both models continue to operate under Anthropic’s broader commitment to safety, with Mythos 5.1’s relaxed controls balanced by the stringent vetting of its user base.

By offering a tiered approach, Anthropic aims to satisfy both mass‑market demand and niche, high‑risk applications without compromising overall system integrity.

Clients interested in Mythos 5.1 must submit detailed use‑case documentation and undergo security assessments before receiving access credentials.

Anthropic’s strategy reflects a broader industry trend of differentiating AI offerings based on risk profile and domain specificity.

Stakeholders in sectors such as finance, healthcare and national security are watching the rollout closely for potential adoption.

Overall, the dual release signals Anthropic’s effort to refine its model portfolio while responding to user feedback on cost and safety.

Why This Matters: The launch of Claude Fable 5.1 and the limited‑access Claude Mythos 5.1 expands affordable, high‑performance AI options for both general developers and specialised, high‑risk organisations.