Users question coding-focused LLM benchmarks

Poster argues new LLM benchmarks and leaderboards skew toward coding while other use cases go untested, noting results can be 'benchmaxxed' without reflecting real quality. Thread draws 60 comments on r/LocalLLaMA.