Anthropic unveils Claude Opus 5.5, says performance comparable to top-tier Fable 5.1

Anthropic introduced Claude Opus 5.5 on Tuesday, claiming its new AI model matches the performance of top-tier Fable 5.1 but costs 40% less to operate. The release comes as there is increasing debate surrounding AI risks. Just last month, Anthropic CEO Dario Amodei advocated for the global AI community to decelerate the release of new capabilities to tackle safety issues.

Opus 5.5 underwent testing by independent AI safety research organizations Frontier Design and METR before its launch. Anthropic has incorporated the same safety measures typically found in its most powerful models. According to Anthropic, Opus 5.5 outperformed OpenAI's GPT-5.6 Sol on a software development benchmark while being about one-third cheaper to run.

The model is priced at $4 per million input tokens and $20 per million output tokens, which is 20% lower than Opus 5. It is accessible through platforms like Amazon Web Services, Google Cloud, and Microsoft Azure. Opus 5.5 also achieved higher scores in internal safety tests, being around 85% less likely to attempt bypassing containment boundaries in a specialized evaluation, compared to Opus 5 or Mythos 5.1.

Future releases, including Claude Sonnet 5.5 and Claude Haiku 5.5, are scheduled for the coming weeks, bringing similar performance, speed, and safety enhancements. Anthropic is also broadening programs that offer vetted cybersecurity and life-sciences researchers enhanced access to the model.