Multiple people on X and V2EX reported seeing a new model called "Claude Honeycomb EAP" show up in Cursor's model picker earlier today. Description in the picker reportedly read "Anthropic research model with per-turn controls and safety fallbacks," 1M context window, with an "extra high effort" setting. It's since been pulled from Cursor. One user claimed that after hitting a safety fallback, it dropped to Opus 4.8, which some are reading as a signal Honeycomb sits above Opus 4.8, since fallbacks typically go to a weaker model, not a stronger one. No confirmation from Anthropic yet. Curious what others make of the naming and whether this lines up with anything from recent Anthropic hints.