evanwtf/local-llm
Which model + engine + agent harness runs a coding agent best on local hardware? Real-repo agent benchmarks, pass rates and wall time, measured on an M5 Max MacBook Pro and a two-node DGX Spark cluster. OpenCode is the primary harness.
GitHub repository with 5 stars and 0 forks.
Language: Python
Topics: agentic-coding, ai-agents, apple-silicon, benchmark, coding-agent, dgx-spark, llama-cpp, llm, llm-inference, local-llm