Anthropic has released Claude Sonnet 5.5, positioning it as a lower-cost, faster complement to Claude Opus 5.5 for everyday coding, bug fixing, and document creation. The closed-weights model is available immediately on the Claude Platform, AWS, Google Cloud, and Microsoft Azure, offering a 1-million-token context window and a 128K max output limit.

List pricing remains identical to Sonnet 5 at $2 per million input tokens and $10 per million output tokens. However, Anthropic highlights that total task costs are lower due to improved token efficiency and fewer required iterations. Early enterprise feedback indicates significant reductions in token usage and task completion times compared to earlier models.

Sonnet 5.5 incorporates adaptive thinking by default across five selectable effort levels. While higher effort levels improve reasoning, Anthropic noted that the highest effort setting can lead to timeouts or penalties on complex benchmarks due to multi-agent code review loops, leaving Opus 5.5 as the preferred choice for open-ended architecture tasks.

Why it matters

  • Identical pricing combined with higher token efficiency reduces net operational costs for enterprise production workloads.

  • Granular effort levels allow developers to optimize API latency and compute costs based on task complexity.

  • Strong performance on terminal tasks accelerates the deployment of automated coding and maintenance workflows.

Source: marktechpost.com