Overview
Anthropic added Claude Sonnet 4.6 as a supported model option in Claude Code, released in version 2.1.45 of the CLI. The change appears in the changelog as a discrete model availability update, expanding the roster of models users can select when running Claude Code sessions.
Previously, Claude Code's default model lineup centered on Opus and earlier Sonnet variants. This is a model availability change, not a pricing or policy update. Sonnet 4.6 was already available via the API; this release brings it into Claude Code's supported model list, making it selectable via the /model picker or --model flag without workarounds.
What this means for brands
For teams using Claude Code in automated pipelines or agentic workflows, Sonnet 4.6 adds a faster, lower-cost option relative to Opus 4.6 while still running on a current-generation model. Brands that have built internal tools or workflows on Claude Code now have more granular control over the cost-quality tradeoff per task. Any brand running evals or benchmarks against Claude Code outputs should account for the fact that the default model may differ across sessions if users or scripts are pinning to 'sonnet' without a version suffix, since version resolution may now point to 4.6.
If your brand surfaces in developer-facing contexts through Claude Code, the model backing those sessions affects how your documentation, APIs, or SDKs are described and cited. Sonnet 4.6 carries the same 128k output ceiling as Opus 4.6 when accessed through Claude Code, which matters for long-form code generation tasks that reference brand assets or developer resources.
What to do
If your team runs Claude Code in any capacity, update to v2.1.45 or later and run your standard query set against Sonnet 4.6 to establish a baseline. Open the /model picker and confirm Sonnet 4.6 appears as an option. If you have scripts or CI jobs calling Claude Code with --model sonnet or a pinned version string, audit those references now: check whether they resolve to 4.6 or an older variant, and update accordingly if you want consistent behavior. If you maintain internal benchmarks comparing Claude Code outputs for brand-relevant tasks, add a Sonnet 4.6 column to your results alongside whatever model you were previously using as the cost-efficient tier.
