Claude Mythos and Fable explained: what Anthropic’s top-tier models mean
Mythos-class is Anthropic’s top tier of Claude models, positioned above Opus. Claude Fable and Claude Mythos are the same underlying model with different safeguards: Fable is generally available with safety classifiers for areas such as cybersecurity and biology, while Mythos has some safeguards lifted and is limited to vetted partners. The current versions, Fable 5.1 and Mythos 5.1, were released in September 2026 and cost $10 per million input tokens and $50 per million output tokens.
What does “Mythos-class” mean?
Anthropic organises Claude models into tiers. For most of 2025 and 2026 the line-up ran from Haiku (fast and cheap) through Sonnet (the everyday default) to Opus (the most capable). Mythos-class is a new tier above Opus, built for long-running, multi-step agentic work such as complex software engineering, research and analysis. Anthropic describes Fable 5.1 and Mythos 5.1 as its most advanced models for coding and knowledge work. Our guide to Opus, Sonnet and Haiku covers the tiers below.
What is the difference between Fable and Mythos?
They are the same model. According to Anthropic’s announcement, Fable 5.1 is generally available, while Mythos 5.1 is available only through trusted access programmes with safeguards designed for work in cybersecurity and the life sciences. Fable uses classifiers that watch for high-risk requests in areas such as cybersecurity, biology and chemistry and model distillation; when a request is flagged, it is answered by a less capable Claude model instead. One UK guide reports this happens in fewer than 5% of sessions.
How did these models come about?
The first Mythos model was announced in April 2026 and withheld from public release because of its cybersecurity capabilities, as CNBC reported. Anthropic shared it with selected software and security companies through Project Glasswing. In June it released Fable 5, a version with extra safeguards that could be offered widely. A brief US export restriction paused access in June before it was lifted at the end of the month, according to Caylent.
What does Fable cost?
Fable is two and a half times the price of Opus 5.5 per token. Anthropic also requires 30-day data retention for traffic on Mythos-class models, which matters for organisations with strict data policies. Our Claude API guide explains how token pricing, caching and batch discounts work.
When should a business use Fable?
- Use Fable when Opus 5.5 has been tested and clearly falls short
- Good candidates: large codebase changes, multi-day research, complex financial or legal analysis
- Avoid Fable for chat, drafting, summarising or anything high-volume
- Measure accuracy and total cost per completed task, not just price per token
Can you use Fable in the Claude apps?
Fable is available through the API and to paid Claude subscribers, with access rolled out in stages because demand is high. Availability and usage limits vary by plan, so check the model picker in your Claude app and your plan’s terms. In coding tools such as Claude Code, you can usually choose the model per task.
Why are the names confusing?
Anthropic has moved from version numbers alone to named tiers, while also releasing frequent point updates. In 2026 alone, teams saw Opus 4.7, Opus 4.8, Fable 5, Mythos 5 and then the 5.5 and 5.1 releases. A simple rule helps: Haiku, Sonnet and Opus are the everyday tiers; Fable is the top tier you can use; Mythos is the restricted twin of Fable.
How UIUX Studio helps
We help teams choose the right Claude model for each task, test cost and quality on real work, and build agents that switch models when needed. Talk to our Agentic AI & Automation team, or read our guide to what Claude is.
Frequently asked questions
Is Claude Fable better than Opus?
Fable is more capable, especially on long agentic tasks, but costs about 2.5 times as much per token.
Can my business use Claude Mythos?
Generally no. Mythos is limited to vetted partners in cybersecurity and the life sciences.
Are Fable and Mythos different models?
No. They share the same underlying model; the difference is the safeguards.
How much does Fable 5.1 cost?
$10 per million input tokens and $50 per million output tokens.
What happens when Fable flags a request?
It is answered by a less capable Claude model instead.