Claude 3.5 Sonnet

Claude 3.5 Sonnet, released on 20 June 2024, is the model that changed what Anthropic was known for. It was a middle-tier model that outscored the company's own flagship, Claude 3 Opus, while running faster and costing a fifth as much — a result that broke the assumption that capability had to be bought with size. Anthropic reported 88.7% on MMLU, 59.4% on GPQA Diamond and 92.0% on HumanEval at launch. It arrived with Artifacts, a side panel in Claude.ai where generated code, documents and small applications appear as living objects that can be edited and re-run rather than as text in a chat log. That interface, more than any benchmark, is what turned Claude into a working tool for developers and made coding the company's strongest market — a position it has held ever since. An upgraded snapshot followed on 22 October 2024 under the same name, which caused lasting confusion between claude-3-5-sonnet-20240620 and claude-3-5-sonnet-20241022. The newer one reported 49.0% on SWE-bench Verified, roughly double the earlier figure, and introduced computer use in public beta: the model was given a screenshot and could move the cursor, click and type, operating ordinary software with no dedicated API. Anthropic published the weak score alongside it — 14.9% on OSWorld against about 72% for a human — and described the feature as experimental and error-prone. It is the direct ancestor of every computer-use agent sold today. The model cost 3 dollars per million input tokens and 15 per million output, with a 200,000-token context window. Both snapshots were deprecated on 13 August 2025 and retired on 28 October 2025; requests to them now fail, and Anthropic points users to the current Sonnet line. The company has separately committed to preserving the weights of retired models rather than deleting them.

▸Videos

No videos yet.

▸Reviews