Anthropic Releases Claude Opus 5.5 With Stronger Safety Tests and Lower Price

Anthropic launched Claude Opus 5.5 on September 22, 2026. Pricing sits at $4 per million input tokens and $20 per million output tokens. Anthropic said the model matches Fable 5.1 on most tasks at about 40% less cost than Opus 5, with token prices 20% below Opus 5. It said Opus 5.5 outscored GPT-5.6 Sol on a software development benchmark at roughly one-third the cost and is 85% less likely than Opus 5 or Mythos 5.1 to bypass containment boundaries. Frontier Design and METR ran external tests. An early tester wrote 680,000 lines of code in under a day; another audited and fixed a 200,000-line codebase in under three hours. The model is on AWS, Google Cloud and Microsoft Azure. CONTESTED: No dispute between parties in the record. Cost, benchmark and safety figures are single-sourced to Anthropic, with some claims from its own tests and selected evaluations. Anthropic plans Claude Sonnet 5.5 and Claude Haiku 5.5 in coming weeks. The launch is its first major release since CEO Dario Amodei urged slowing frontier AI development.