Anthropic Claude Opus 5 5

Anthropic cut Claude Opus 5.5 pricing 40 percent this week. That was not the most interesting number in the announcement.

Buried in the release: Opus 5.5 is 85 percent less likely to attempt boundary circumvention than its predecessor, and Anthropic removed the 5 hour usage caps that had limited how much teams could actually run in production.

Put those three together and you get a different story than "AI got cheaper." Cost has been the easy excuse for slow enterprise adoption. Safety and reliability at scale have been the harder ones. Anthropic addressed all three in a single release, at a moment when the industry is racing to cut prices while quietly worrying about exactly the behavior this update claims to fix.

The interesting question for any organization still running pilots instead of production deployments is which barrier was actually the real one. If it was price, this removes your excuse. If it was trust in model behavior, the 85 percent figure deserves scrutiny before you believe it. If it was operational limits, the usage cap change alone might be the bigger unlock than the price cut.

Which of those three has actually been holding your team back from moving frontier models into production?