Anthropic Claude 3.5 Sonnet Review — Computer Use, SWE-bench 49%, and the Model That Redefined the Sonnet Tier

Claude 3.5 Sonnet (June and October 2024) was the first Sonnet-tier model to outperform its predecessor's Opus on benchmarks — and the first publicly available model with computer use capability. At launch it set the SWE-bench record at 49.0%, unseating GPT-4o. The October 2024 update brought computer use (GUI automation), improved coding and instruction-following, and cemented Claude 3.5 Sonnet as the dominant coding LLM for most of late 2024.