Anthropic has released its new flagship model, Claude Opus 5.5. The company claims the model outperforms OpenAI’s GPT-5.6 Sol on software development benchmarks while operating at approximately one-third of the cost. Typical workloads are reported to be 40% cheaper to run compared to the previous Opus 5 version.
The pricing for Opus 5.5 is set at $4 per million input tokens and $20 per million output tokens, representing a 20% reduction from its predecessor. Cache read costs have been lowered by 60% to $0.20 per million tokens. Additionally, the model generates output more than 30% faster than the previous generation.
Before the launch, the model underwent safety evaluations by external groups, including Frontier Design and METR. Anthropic reports that Opus 5.5 achieved its highest scores to date on an automated behavioral audit. The company states the model is 85% less likely to attempt to bypass containment boundaries than Opus 5 or Mythos 5.1.
Anthropic plans to release Claude Sonnet 5.5 and Claude Haiku 5.5 in the coming weeks. The company is also expanding access to the model's biology and cybersecurity capabilities for verified researchers and practitioners through its verification program.
