Anthropic released Claude Sonnet 4.6, a significant update focused on code assistance, longer contexts, agent planning and analytics.

Key improvements

The model introduces a beta context window of 1 000 000 tokens, enabling processing of entire codebases, large contracts or multiple studies in one request.

Developers report improved code quality, more accurate instruction following and a reduction in hallucinations and duplicated logical blocks during generation.

Anthropic also enhanced the "computer use" capability, permitting the model to operate software workflows without dedicated APIs or external connectors for many tasks.

Availability and pricing

Sonnet 4.6 is now the default model on claude.ai for both Free and Pro plans, replacing the previous Sonnet 4.5 baseline.

Pricing remains at the Sonnet 4.5 level, starting from $3 / $15 per million tokens depending on the plan and usage tier.

Implications for users

The increased context window reduces the need for external retrievals or chunking when working with extensive technical material and long-form documents.

Improvements in agent planning and analytics aim to support more complex automation scenarios and multi-step workflows managed directly by the model.