Claude 3.5 Sonnet is Anthropic's most capable model for everyday use, setting benchmark records on coding (HumanEval 93.7%) and graduate-level reasoning at launch. It is the first model to offer Anthropic's "computer use" feature — enabling agentic workflows that interact with GUIs directly.
Key capabilities
- Best-in-class coding — top scores on HumanEval, SWE-bench and SWE-bench Verified
- Computer use (beta) — can control a computer interface to complete multi-step tasks
- 200K context window — fits very long documents and codebases
- Vision — interprets screenshots, charts, and photographs
- Fast and efficient — faster than Claude 3 Opus at lower cost
What it's best for
Coding agents, software engineering tasks, instruction-following pipelines, document analysis, and computer-use automation. The primary model powering Claude.ai's paid plans.