Vals-Smith uses your code library test model: who's better suited for the AI assessment platform Vals to launch Vals-Smith

Vals launched the AI evaluation platform, Vals-Smith, which is able to access the GitHub code library, extract real development tasks from the merged PR and generate an exclusive programming evaluation. Each Question contains a description of the problem, a hidden test and a recyclable environment, and the detected model can only see the pre-repair code and cannot view the original patch. Teams can get different models and coding parties to run the same set of questions, directly comparing pass rates. Vals also completed the $40 million A round of financing, valued at $400 million, with a 16z-funded project. The results of the platform have been cited in model cards from OpenAI, Anthropic, Google, Meta and XAI, and this year income has been eight times higher than in 2025。