Poor Paul's Benchmark as an MCP server. Queryable GPU inference data — quantization, throughput, VRAM, concurrent users — for Claude Desktop, Cursor, Windsurf, Cline, and any MCP client. Self-host or use the free hosted endpoint.
inference quantization gpu-benchmark huggingface llm local-llm model-context-protocol fastmcp poor-pauls-benchmark
-
Updated
May 11, 2026 - Python