Scale AI ran the Claude models with default configurations: high reasoning effort, a 200k-token context window, and a 32k-token limit per response.The freelance-work automation evaluation by CAIS and Scale AI: 240 real freelance projects (game development, product design, architecture, data analysis, and more); human evaluators compare each agent deliverable with a human professional's, and the automation rate is the share judged at least as good as the professional's.