Anthropic’s reported Fable 5.2 and limited Mythos 5.2 releases could broaden multi-agent and cyber-relevant capabilities, but the discovery-class report leaves key performance and safety evidence unverified.
Limits of the evidence: This is a discovery-class, single-source roundup. It attributes capability, pricing, and safety-review statements to Anthropic, and gives no published results for the sandbox-escape tests.
What to watch
Assessment revised Nov 12, 2026
Anthropic releases new top models: Meta’s Muse Ember rivals leading fast models
Andrew NgNov 11, 2026First report
First assessment, from Andrew Ng
No maturity ladder applies to this desk.
Sens.ai aggregates and assesses published reporting. The assessment above is machine generated; the original sources are authoritative.