Community suggests local models for Claude Code agents

A Reddit user proposes Anthropic release optional local models to offload compute for agent tasks in Claude Code, leveraging users' beefy systems to save on tokens. The idea suggests specialized 8b or 30b parameter models running in parallel to maintain quality while reducing API costs.