Anthropic launched two models on 9 June: Claude Fable 5 and Claude Mythos 5. Same underlying model, different leash. Here is what founders actually need to know.
Fable 5 is a "Mythos-class" model, a new tier that sits above the Opus line. It is state of the art on nearly every benchmark Anthropic tested, and the gap widens the longer and more complex the task gets.
Stripe ran an early test where Fable performed a codebase-wide migration on a 50-million-line Ruby codebase in a day. By hand, that was a two-month job for a whole team.
The clever bit nobody else is doing
Releasing something this capable is a liability, so Anthropic built in a fallback rather than a refusal. When Fable detects a query touching cybersecurity, biology and chemistry, or model distillation, the response gets handled by Opus 4.8 instead, and the user gets told.
The classifiers are tuned conservatively, so they will catch some harmless requests, but Anthropic says fewer than 5% of sessions trigger them at all. For everyone else, Fable performs identically to the unrestricted Mythos 5.
That is the actual product story here. Most labs ship a model and bolt on refusals. Anthropic shipped a capability ceiling with a graceful step-down, so a flagged request still gets a strong answer rather than a wall.
Mythos 5 is the same model with the safeguards off
Mythos 5 is Fable with the cyber guardrails lifted, and it is going only to a small group of cyber defenders and infrastructure providers through Project Glasswing, in collaboration with the US government.
Anthropic claims it has the strongest cybersecurity capabilities of any model in the world. A biology trusted-access program is coming for select researchers.
Knowledge work is the part most founders will care about. Fable posted the highest score on Hebbia's finance benchmark for senior-level reasoning, with real gains in document reasoning and chart and table interpretation.
It is also the new top model for vision tasks, including rebuilding a web app's source code from screenshots alone. On the science side, Anthropic's internal protein-design team reported Mythos accelerating parts of drug design roughly tenfold.
The pricing and the catch
Both models cost $10 per million input tokens and $50 per million output tokens, less than half what Mythos Preview cost. But the subscription rollout is where founders should pay attention.
Fable 5 is included free on Pro, Max, Team, and seat-based Enterprise plans from launch through 22 June. On 23 June it gets pulled from those plans, and using it after that needs usage credits.
Anthropic says it will restore Fable as a standard plan feature once capacity allows, but there is no firm date. API and consumption-based Enterprise plans get it fully from day one.
So if you are on a subscription and want to stress-test this thing, you have a two-week window before the meter starts running.
Business customer data on Mythos-class models now carries a mandatory 30-day retention policy across first and third-party surfaces. Anthropic says it will not train on that data or use it for anything beyond safety, with logged access and deletion after 30 days. Worth knowing if your data governance is tight.
