Claude 3.5 Sonnet

Claude 3.5 Sonnet marked a pivotal step in the Claude family, arriving as the first release in the 3.5 lineup before Haiku and Opus completed the series. This model introduced the Artifacts feature—a capability that lets users instantly view and interact with generated code, documents, and other structured outputs rather than receiving them solely as text. Early comparisons positioned it against GPT-4o and Gemini 1.5 Pro, where it demonstrated competitive or superior performance across a range of evaluations, including graduate-level reasoning, undergraduate knowledge, and coding tasks.

The model achieved strong scores on benchmarks like BIG-Bench Hard and GSM8k, reflecting strong multi-step reasoning and mathematical problem-solving abilities. Its HumanEval performance indicated solid functional correctness for code synthesis. As a mid-tier offering within the Claude lineup, it delivered frontier-level intelligence at improved speed relative to heavier models. After its initial release, subsequent family members—Claude Sonnet 5 achieving 82.1% on SWE-Bench and Opus 4.6 introducing agent team features—built on this foundation, expanding the model family's capabilities for enterprise and developer use cases.