02 / IT · Analysis
Anthropic models differ in research request restrictions
Reportedly identical core capabilities do not mean the two models allow the same biology and cybersecurity requests.

DigitalToday, a Korean technology news outlet, reports that Anthropic has released Fable5.2 and Mythos5.2; the supplied report does not specify their licence or where they run.
According to DigitalToday, both models improve performance in scientific research and coding. That description concerns capability; it does not establish which research requests the models accept.
The report describes their core capabilities as identical but distinguishes their safeguards in biology and cybersecurity. The difference in request restrictions therefore matters alongside the reported performance improvements.
Restrictions on the general-use model
DigitalToday describes Fable5.2 as the model offered to general users. It reports that the model blocks most questions related to biological research.
The same report says Fable5.2 also blocks many cybersecurity requests. Improved research performance alone does not imply that a request in either field will be accepted.
The scope of relaxed restrictions
DigitalToday describes Mythos5.2 as a version with those restrictions relaxed. That description does not establish that every research request is permitted.
The distinction presented in the report concerns the models’ safeguards. Choosing between them for research therefore requires considering both capability and the scope of permitted requests; stronger performance alone does not resolve that choice.