smevals: A Small Eval Suite for Models, Prompts, and Harnesses

Simon Willison published smevals, a small evaluation suite intended to assess models, prompts, and harnesses, meaning the surrounding execution or agent framework. The supplied source contains no abstract or detailed results, so the available facts are limited to the project’s stated scope and title. Its task design, metrics, supported models, implementation, licensing, and comparative results require verification from the original post and repository before drawing stronger conclusions.