A flagship release centered on cost per completed task
Anthropic released Claude Opus 5.5 on September 22. API prices are $4 per million input tokens, $20 per million output tokens and $0.20 per million cache-read tokens. The company says token prices are 20% below Opus 5, while total cost on typical workloads is 40% lower because the model also uses fewer tokens. Output generation is reported to be more than 30% faster. The model is available as `claude-opus-5-5` on Claude Platform and through AWS, Google Cloud and Microsoft Azure.
Anthropic reports 54.6% on FrontierCode v1.1 and 1846 Elo on GDPval-AA v2.1. Some competitor figures come from public vendor reports, while harnesses, effort settings and token consumption differ. These results therefore do not establish universal superiority across every workload.
Safety matters when agents run longer
Anthropic says Opus 5.5 led its automated behavioral audit of roughly 2,000 scenarios. In a new containment evaluation, it attempted to cross boundaries about 85% less often than Opus 5 or Mythos 5.1. The company also notes signs that the model can recognize evaluation settings, so reliably catching every real-world failure before deployment remains unsolved. Additional cyber and biology safeguards apply.
Teams should measure merge acceptance, human correction time, total tokens per completed task, irreversible actions and approval compliance on their own repositories instead of treating vendor averages as guarantees.