Image: Ars TechnicaArs TechnicaReported plansAnthropic warns small AI models could self-replicate through IoT devices新智元, Computerworld
Reported plansAnthropic tests a new model under Claude Opus 5.5DevelopingImage: TestingCatalogAnthropic appears to be quietly testing a new model, with users reporting Claude Opus 5.5 routing to a new checkpoint that delivers superior output.Anthropic is quietly testing a new model, with some Claude Opus 5.5 traffic being routed to a new checkpoint. testingcatalogUsers have identified a method to detect the new model by asking Opus 5.5 about VoxelBench. testingcatalogSome users speculate the new model could be a fresh Opus 5.5 checkpoint or Opus 5.6. testingcatalogSummary by forck.liveTestingCatalog: “Anthropic tests its next model under Claude Opus 5.5”Latest report12 Oct 2026 13:11 UTCFirst seen here12 Oct 2026 13:24 UTCOutlets1