Came across a site today, https://hnup.date/hn-sota. Unlike most leaderboards that focus on benchmark scores, this one primarily pulls authentic comments from top Hacker News developers and runs AI sentiment analysis on them to evaluate models.
Based on what it shows, Claude (Anthropic) remains the strongest model in developers’ eyes — recognized for a superior “engineering sensibility” in complex logical reasoning and code refactoring that’s currently considered top-tier. GPT (OpenAI) comes in second, keeping an edge in generation speed, ecosystem support, and reliable standard implementations across mainstream languages. And DeepSeek stands out as an impressive dark horse, earning a strong practical reputation in hardcore tech circles through extreme cost-effectiveness and deep language-understanding capability.
