Theo Browne says Grok 4.7 “probably should not be used,” sharing a cost-versus-intelligence chart. You thought you were choosing a model; apparently you were auditioning for a developer’s cancellation announcement. In a follow-up, he says people who called it the best model at launch should not be trusted. That is Theo’s judgment, not a recall notice or our universal performance verdict. The replies are not unanimous: erik defends its X-connected research, while Keith Crabb says he still likes it for audits. We haven’t independently verified the shared chart’s underlying runs. Different tasks can reward different strengths. For now, the model has not been discontinued. It has merely been escorted out of Theo’s personal leaderboard, along with the people who applauded too loudly.