Anthropic released Claude Opus 4, which the company demonstrated executing code and task sequences for up to 7 hours (with customer Rakuten) and game-play for 24 hours, compared to 45 minutes for the prior Claude 3.7 Sonnet model. This represents a material extension in agentic runtime and task continuity without human intervention. The improvement directly tests whether multi-step completion reliability is advancing at speeds and scales that could translate into enterprise adoption and ROI for knowledge-work automation. For financial institutions, this data point bears on whether agentic reliability gains are creating durable use cases in software development and business process automation.
238 other tracked developments touching the same claims.