Y Combinator’s AI Office Hour Simulator reportedly switches to GLM-5.2 on Wafer

Wafer says Y Combinator built AI versions of its partners to help people talk through startup ideas. According to Wafer, YC tested lightweight Gemini and OpenAI models before moving its Office Hour Simulator to GLM-5.2 on a dedicated Wafer endpoint. YC then compared answer quality, latency and conversation duration against GPT-4.1 mini on OpenAI and Gemma 4 31B on Cerebras. Wafer reports 31% and 44% lower average model latency, respectively, and says users spent 2.5 minutes longer talking to YC’s AI partners when using Wafer.