Developers Share Favorite AI Models for Everyday Features

A recent Hacker News thread asked developers what AI models power the everyday features in their products—things like generating content, rewriting text, recommendations, and contextual suggestions, rather than coding assistants or agentic workflows. The responses highlight a practical, cost-conscious approach: many developers prioritize affordability, speed, and reliability over chasing the latest frontier model. The thread shows that for structured workloads, models like Google's Gemini Flash (often accessed via OpenRouter) are popular choices due to their low latency and sufficient quality. However, developers also noted the challenge of navigating dozens of available models, with many building custom evaluation suites to compare cost, latency, and performance rigorously. Others rely on a handful of trusted models rather than testing the entire landscape. The discussion underscores that for most user-facing AI features, a less capable but faster and cheaper model often wins out over the most advanced one.