Anthropic Launches Cheaper Sonnet 5.5 Days After CEO Called for AI Industry to Slow Development Pace

Sonnet 5.5 gives buyers a practical choice about what they are paying for. Anthropic positions it for routine execution, reserving Opus for work requiring more reasoning and judgment. Matching the model to the task is a more useful purchasing question than always choosing the most capable option.

Tekedia reports pricing of USD 2 per million input tokens and USD 10 per million output tokens, half Opus 5.5’s rates. Anthropic also says Sonnet needs fewer tokens than its predecessor. The price is a stated rate; the efficiency improvement is a company claim. Buyers still need to assess total cost on their own workloads, including whether the results meet their requirements.

The timing need not contradict CEO Dario Amodei’s call to slow frontier development. Anthropic says Sonnet does not advance its capability frontier. Selling a cheaper option and pursuing a more powerful frontier model are different activities.

That distinction is not a safety exemption. The company also reports stronger cybersecurity capabilities than the previous Sonnet and has added safeguards similar to those used for its most capable models. Better capability and better protection are separate claims. The release deserves assessment on both, rather than assuming a lower price makes its risks smaller.