Claude Fable 5 or Mythos 5: which should you use?

A week after Anthropic launched its two Mythos-class models, Claude Fable 5 and Claude Mythos 5, a common question is a practical one: which should I pick? The short answer is that most people cannot pick Mythos 5 at all. It is the same underlying model as Fable 5 with some safeguards lifted, and Anthropic offers it only to Project Glasswing partners, its group of cyber defenders and critical software providers, and soon to selected biology researchers through a trusted access program.

So for an individual, a small business or a typical developer, the choice is really between Fable 5, Anthropic's most capable generally available model, and Claude Opus 4.8, the model released in late May that now sits one step below it.

Who can use what

If you work in cybersecurity and your organization does not have Mythos Preview access today, watch for the more systematic application process Anthropic says it is planning. Until then, Fable 5 is your option, and its cyber classifiers will pass many security requests to Opus 4.8.

Matching the model to the task

Anthropic's guidance gives three useful signals. Fable 5's advantage over other Claude models grows as tasks get longer and more complex. Its classifiers send requests about cybersecurity, biology and chemistry, or attempts to copy the model, to Opus 4.8, and for now that covers most biology and chemistry questions. And in Anthropic's early data, over 95% of Fable sessions never needed that hand-off. Dianne Na Penn, who leads product management for research and labs at Anthropic, told Fortune that customers should bring Fable 5 their hardest work.

Here is how that plays out for typical tasks:

A simple way to decide on your own work is to run one real task on both models with the same prompt:

Here are 40 pages of meeting notes from the last three months.
Produce a decisions log: one row per decision, with the date,
who decided, what was decided and any open follow-up.
Before you finish, check each row against the notes and list
any decision you were unsure about.

Compare the two logs for missed decisions and wrong dates, not for style. If Opus 4.8 gets them all, the cheaper model is the right one for that job.

A note on data for business users

Anthropic now requires 30-day retention of all business traffic on Mythos-class models, which includes Fable 5. It says the data will not be used for training, that human access is logged, and that it will be deleted after 30 days in almost all cases. If your organization has strict rules about where data goes, check them before moving sensitive work to Fable 5.

What does each cost?

On the Claude API, Fable 5 and Mythos 5 cost the same. Prices are quoted per million tokens, the small pieces of text a model reads and writes, often parts of words, with separate rates for what you send (input) and what the model writes back (output).

Opus 4.8 is half the price of Fable 5. Its fast mode, which Anthropic says runs 2.5 times faster, costs the same as Fable 5.

A worked example makes the difference concrete. Suppose a report job sends 200,000 tokens of notes and gets back 20,000 tokens of output:

The token counts are illustrative; the prices are Anthropic's published rates. For one report, the difference is small. For an app that runs the same job a thousand times a day, it is the difference between $3,000 and $1,500 daily, so test whether the cheaper model is good enough before you scale.

On subscription plans, the timing matters. Pro, Max and Team subscribers, and Enterprise customers billed per seat, get Fable 5 at no extra cost through Monday, June 22. From June 23, using it on those plans will require usage credits, unless Anthropic has the capacity to extend the included period, which it says it will try to do. If you have a large Fable 5 task in mind, this week is the time to run it.