Claude Sonnet 4

Claude Sonnet 4 marks a deliberate step into agentic AI workflows, built as a capable everyday model that sits beneath the flagship Opus tier while still delivering meaningful improvements over prior generations. The release introduced extended thinking with tool use, allowing the model to interleave reasoning with live tool calls, and to use multiple tools in parallel during a single response. Developers gained significantly improved instruction adherence, enabling tighter control over how the model interprets complex or layered requirements, along with a feature that lets Claude save facts from local file interactions over time, building a form of persistent project memory. These capabilities collectively position Sonnet 4 as a practical backbone for coding assistants, internal copilots, and any workflow that requires sustained, multi-turn reliability under nuanced instructions.

The subsequent 4.6 release sharpened the model's focus on software engineering, bringing measurable gains in coding precision and PC usage performance alongside expanded developer tools and API enhancements. Independent tests showed Sonnet 4.6 outperforming comparable models from other major providers across multiple benchmarks, and the 1 the cataloged API limit, expanded from earlier generations, gives the model room to work with large repositories and multi-document tasks without losing coherence. Early adoption in GitHub Copilot highlighted the model's strength in agentic coding scenarios, where it handles multi-step tasks that require planning, tool use, and error recovery over extended interactions. For teams seeking broad AI capability without the cost and latency profile of top-tier reasoning models, Sonnet 4 occupies a practical and capable middle ground.