Anthropic releases Claude Opus 5.5 with performance matching top-tier models at 40% lower cost, emphasizing safety improvements.
Anthropic unveiled Claude Opus 5.5, a new AI model delivering performance on par with its top-tier predecessor while cutting operating costs by 40 percent. The release comes as the company emphasizes safety measures, with the model undergoing external testing before launch and showing improved safeguards.
- ▪Opus 5.5 costs $4 per million input tokens and $20 per million output tokens—20% cheaper than Opus 5.
- ▪The model outperformed OpenAI’s GPT-5.6 Sol on software development benchmarks while costing roughly one-third as much to operate.
- ▪Independent safety groups Frontier Design and METR conducted external testing before the model’s release.
- ▪Opus 5.5 was 85% less likely than predecessor models to attempt bypassing containment boundaries in dedicated safety evaluations.
- ▪Claude Sonnet 5.5 and Claude Haiku 5.5 versions launching in coming weeks with similar performance and safety improvements.
- ▪Available on Amazon Web Services, Google Cloud, and Microsoft Azure.
- ▪Anthropic CEO Dario Amodei recently called for the AI community to slow development pace to address safety concerns.
- ▪The company is expanding access programs for vetted cybersecurity and life-sciences researchers.
Source: according to media reports
COMMENTS