Anthropic Debuts Claude Opus 5.5 as First Model in Claude 5.5 Line, Cutting Costs and Raising Performance

According to AI TIMES, Anthropic unveiled its next-generation model, Claude Opus 5.5, on September 22, saying it cut operating costs by 40% from the previous version while improving performance on coding and specialized knowledge tasks. In early tests, the company highlighted a case in which it completed a 680,000-line code migration within a day, emphasizing that the model is designed for long-running development work and agent operations. Anthropic said its own evaluations showed performance gains across a range of real-world task tests, including coding benchmarks, enterprise analysis tasks, financial work, and web optimization. On safety, it said the model underwent external evaluation and internal audits, and that it strengthened controls for prompt-injection defense and long-duration autonomous execution. The article noted, however, that many of these performance and cost figures are based on the company's own evaluations and early cases, and may vary in real enterprise environments depending on workload type and system configuration. Anthropic plans to make the model available on its own platform and major clouds, and to release additional models in the same family.