Anthropic Releases Claude Fable 5, The Mythos Model the Public Can Finally Use

Two months after Anthropic unveiled the Mythos to a limited number of users, citing concerns about its potential misuse, the company is releasing an equally powerful model to the public.

Today, it launched Claude Fable 5, a Mythos-class model made safe for general use. Fable 5 exceeds the capabilities of any model Anthropic has previously made generally available.

The name is deliberate: Fable derives from the Latin fabula, meaning “that which is told,” and shares an intentional connection with the Greek mythos.

The pairing positions Fable as the public counterpart to the restricted Mythos. Fable’s launch comes as Anthropic exceeds a projected $900 billion valuation and prepares to enter the public markets alongside OpenAI and Elon Musk’s SpaceX.

What Fable 5 Can Do and Where It Is Blocked

Fable 5 is state-of-the-art across nearly all tested AI benchmarks, excelling in software engineering, knowledge work, vision, scientific research, and more. 

Its advantage grows with task complexity, allowing it to work autonomously for longer than any previous Claude model, including multi-step coding, extended research, and tool-based workflows. 

The gap over the Opus 4.8 is largest in long-horizon software engineering, large-scale document understanding, and agentic tool use.

In high-risk areas such as cybersecurity, biology, chemistry, and distillation, Fable 5 blocks responses and falls back to Claude Opus 4.8. 

As Anthropic confirmed, this fallback makes public release possible, placing Mythos-level capabilities behind a safety layer that cannot be prompted, jailbroken, or overridden. 

Anthropic says the safeguards activate in less than 5% of sessions, making the fallback invisible for most users. 

Mythos 5, the Pricing Structure, and the Scientific Research Angle

Beyond Claude Fable 5, Anthropic also launched Claude Mythos 5, which is a Fable 5 version with fewer safeguards reserved for cyberdefence and infrastructure providers that already had Mythos Preview access through Project Glasswing. 

Both models cost $10 per million input tokens and $50 per million output tokens, less than half the Mythos Preview price, and include a 90% discount on cached prompts, making long-context enterprise workflows far more affordable.

Anthropic’s most striking claim centers on scientific research. The company used the unrestricted model to assist Dyno Therapeutics in developing new adeno-associated virus capsid candidates for gene therapy. 

Dyno’s team confirmed that the model identified candidates that performed as well as or better than the best results from its human-guided search process. 

If confirmed in further trials, the result would be one of the clearest examples of frontier AI delivering genuine scientific value beyond productivity gains.

Access, the IPO Context, and the Recursive Self-Improvement Warning

Last week, Anthropic expanded Mythos access to hundreds of organisations across 15 countries, focusing on critical infrastructure operators. 

The launch also follows Anthropic’s call for major AI labs to establish a coordinated brake on frontier AI development, warning that systems may soon reach recursive self-improvement, autonomously improving themselves without human intervention.

That tension has not gone unnoticed. Anthropic is releasing a model it spent two months describing as too dangerous for general use while warning that frontier AI is approaching a threshold where today’s safety decisions could become irreversible. 

With Claude Fable 5, Anthropic is advancing its stated goal of deploying Mythos-class models at scale while capitalising on growing investor interest ahead of a potentially massive IPO. 

Source: Claude Fable 5 and Claude Mythos 5