TL;DR: Anthropic has launched Claude Fable 5 for general enterprise deployment and Claude Mythos 5 for specialized cybersecurity applications. Both models cost $10 per million input tokens and $50 per million output tokens, offering state-of-the-art capabilities in software engineering, vision, and long-context processing. See our Full Guide to learn how to prepare your development workflows for these models.

In early 2026, enterprise developers can deploy Anthropic’s newly released Claude Fable 5 and Claude Mythos 5 models to scale automated software development and analytical workflows. Following a temporary service disruption, Anthropic structured this dual release to offer its most powerful cognitive architecture under two distinct safety profiles. Fable 5 provides safe, general-use cognitive performance for enterprise applications, while Mythos 5 delivers unrestricted access to cybersecurity and infrastructure defense partners.

How Do Fable 5 and Mythos 5 Differ in Safety and Access?

Claude Fable 5 contains automated safety safeguards that redirect sensitive queries to Claude Opus 4.8, whereas Claude Mythos 5 operates with these restrictions lifted in specific technical domains. To release Fable 5 quickly, Anthropic tuned these safeguards conservatively. They trigger on average in less than 5% of sessions, occasionally routing harmless requests to Opus 4.8. This guardrail prevents the misuse of Fable 5's highly capable code generation and analysis in sensitive areas like cybersecurity.

Understanding the Project Glasswing Deployment

Mythos 5 is the same underlying model as Fable 5 but lacks these active safeguards. Initially, Anthropic is deploying Mythos 5 through Project Glasswing, an initiative run in collaboration with the United States government. This framework is an upgrade to Claude Mythos Preview, giving cyberdefenders advanced tools to secure critical software systems. Anthropic plans to expand access to Mythos 5 soon through a broader trusted access program.

Fable 5 Delivers State-of-the-Art Software Engineering Automation

Fable 5 automates complex, codebase-wide tasks over longer operating periods than any previous Claude model. During early testing, Stripe used the model to compress months of work into days, migrating a 50-million-line Ruby codebase in a single day. On Cognition's FrontierCode evaluation, which measures a model's performance on difficult coding challenges while meeting production standards, Fable 5 achieved the highest score among all frontier models at a medium-effort setting.

Code Migration and Token Efficiency

Fable 5 is highly token-efficient, allowing developers to execute extensive refactoring tasks without exhausting context limits. The model works autonomously for hours, reading documentation, planning migrations, and verifying outputs. This capability reduces the engineering hours required to maintain legacy codebases and integrate new APIs.

How Much Do Fable 5 and Mythos 5 Cost to Run?

Fable 5 and Mythos 5 cost $10 per million input tokens and $50 per million output tokens. This flat rate represents a price reduction of more than 50% compared to the older Claude Mythos Preview model. This pricing structure lowers the operational cost of deploying autonomous agents for large-scale enterprise workflows.

Memory Retention and Long-Context ROI

Fable 5 maintains focus across millions of tokens and uses its own file-based notes to refine its outputs over time. During testing with the deck-building game Slay the Spire, giving the model access to persistent file-based memory improved its performance threefold compared to Opus 4.8. Fable 5 also reached the game’s final act three times more often, demonstrating that persistent memory significantly increases success rates in multi-stage business operations.

Advanced Vision and Analytical Performance Set New Benchmarks

Fable 5 outperforms previous models on visual processing, financial reasoning, and scientific analysis tasks. The model extracts precise numerical values from dense scientific charts and can rebuild functional web application source code directly from static screenshots. It executes these visual tasks with minimal external scaffolding, demonstrating a high degree of independent spatial reasoning.

Financial Analysis and Drug Discovery Benchmarks

In analytical tasks, Fable 5 scored higher than all other tested models on Hebbia's Finance Benchmark for senior-level reasoning. The trading firm IMC reported that Fable 5 successfully completed its trading-analysis evaluations, including conceptual reasoning, factual lookup, and expected-value calculations. In life sciences, internal experts used Mythos 5 to accelerate protein design workflows by approximately ten times without human assistance, illustrating its potential to speed up therapeutics development.

Key Takeaways

  • Fable 5 and Mythos 5 offer a 50% cost reduction compared to Claude Mythos Preview, pricing input tokens at $10/M and output tokens at $50/M.
  • Fable 5 features conservative safeguards that redirect less than 5% of queries to Opus 4.8, while Mythos 5 provides unrestricted access for cybersecurity professionals via Project Glasswing.
  • The new models show high token efficiency and strong autonomous performance, enabling rapid codebase migrations and highly accelerated drug discovery workflows.