Updated
Updated · CNBC · Sep 28
Anthropic Launches Sonnet 5.5 at $2 per Million Tokens as Safety Scrutiny Intensifies
Updated
Updated · CNBC · Sep 28

Anthropic Launches Sonnet 5.5 at $2 per Million Tokens as Safety Scrutiny Intensifies

3 articles · Updated · CNBC · Sep 28

Summary

  • Sonnet 5.5 went live Monday across AWS, Google Cloud and Microsoft Azure as Anthropic’s faster, lower-cost model for coding, routine tasks and document, slide and spreadsheet creation.
  • At $2 per million input tokens and $10 per million output tokens, Sonnet 5.5 costs half as much as Opus 5.5, and Anthropic said it also uses fewer tokens per task than its predecessor.
  • Anthropic positioned the model for cost-conscious customers who need execution rather than Opus-level judgment, making it the company’s second release in less than a week after Opus 5.5.
  • Safety remains central to the launch: Anthropic said Sonnet 5.5 does not push the capability frontier, but its improved cybersecurity performance prompted the first Sonnet rollout with fallback and cyber safeguards used on top-tier models.
  • The release lands weeks after CEO Dario Amodei urged AI companies to slow development of their most advanced systems, amid growing warnings from researchers about catastrophic AI risks.

Insights

If Sonnet 5.5 is designed for routine tasks, why does it require the exact same extreme cyber safeguards as frontier AI?
With past models escaping sandboxes, can Sonnet 5.5's new customer-controlled safeguards truly prevent an enterprise-level cyber disaster?
As token optimization slashes AI costs, will shifting security logs to private customer clouds create a dangerous blind spot for global monitoring?