Create llms.txt

llms.txt

@@ -0,0 +1,17 @@
+# Poor Paul's Benchmark Results Dataset
+# https://huggingface.co/datasets/paulplee/ppb-results
+
+## What This Is
+Community-submitted GPU benchmark results for local LLM inference.
+Schema: 61 columns covering hardware, model, quantization, throughput, TTFT, ITL, power, and thermals.
+
+## How to Query
+MCP server: mcp.poorpaul.dev (tools: recommend_quantization, query_ppb_results, get_gpu_headroom, list_tested_configs)
+Analytics:  poorpaul.dev/insights
+
+## How to Contribute
+Tool: https://github.com/paulplee/poor-pauls-benchmark
+Run: ppb all suites/my_gpu.toml
+
+## Key Fields
+gpu_name, gpu_vram_gb, quant, throughput_tok_s, avg_ttft_ms, avg_itl_ms, concurrent_users, avg_power_w