Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?

A Hacker News discussion asking whether developers have successfully replaced Claude or GPT with local language models for daily coding work. The post seeks real-world setups, performance metrics, and practical experiences from developers who have made this transition, rather than just experimental usage.

Key Points

• •Evaluate local models as production-grade replacements for Claude/GPT in daily coding workflows
• •Consider performance metrics like tokens per second (tok/s) when benchmarking local alternatives
• •Assess trade-offs between API-based models (Claude/GPT) and self-hosted local models
• •Document complete setup configurations for reproducible local coding environments
• •Measure real-world coding productivity and code quality with local models
• •Identify which coding tasks benefit most from local model deployment
• •Compare latency, cost, and resource requirements of local vs. cloud-based solutions
• •Share hardware requirements and infrastructure needed for viable local model performance

Workflow Diagram

Concepts