Community benchmark suites for evaluating local LLM quality. Submit results via the API.
A five-prompt creative writing eval where models draft short tech-related 4chan-style greenposts. DeepSeek judges format compliance, reasonable length, tech relevance, coherence, and humor.