Benchmark local LLM models — speed, quality & hardware fitness verdict from any

{
  "mcpServers": {
    "metrillm": {
      "args": [
        "-y",
        "metrillm-mcp"
      ],
      "command": "npx",
      "transport": "stdio"
    }
  }
}