10 actions in 10 pieces of software · ranked by words, meaning and measured trust · 639 ms

1. 1jetson-llm-benchmarkverified and signedpassport

   skillnvidia/skillsrecord

   • jetson-llm-benchmark

     Benchmark Jetson LLM/VLM serving performance across vLLM, llama.cpp, and Ollama with structured JSON output.

2. 2benchmark-vllm-modelverified and signedpassport

   skillchalk-ai/chalk-ai-pluginsrecord

   • benchmark-vllm-model

     Benchmark throughput and cost-per-token of a model served on vLLM in Chalk's container infrastructure. Use when someone asks for $/token, $/request, tokens/sec, "how much would it cost to self-host this model", GPU sizing for an LLM, or wants to compare a self-hosted model's economics against a host

3. 3jetson-inference-mem-tuneverified and signedpassport

   skillnvidia/skillsrecord

   • jetson-inference-mem-tune

     Pick the serving stack and per-runtime memory flags (vLLM, SGLang, llama.cpp, TensorRT Edge-LLM) for an LLM/VLM workload on any NVIDIA Jetson.

4. 4LLM MCP Bridgefound in code, not yet checkedno passport yet

   MCP serverramgeart/llm-mcp-bridge

   • llm_benchmark

     Ejecuta un benchmark con múltiples prompts para evaluar rendimiento del modelo

5. 5jetson-llm-serveverified and signedpassport

   skillnvidia/skillsrecord

   • jetson-llm-serve

     Stand up vLLM or SGLang serving on Jetson, using upstream vLLM on Thor and Orin JetPack 7.2+, and NVIDIA-AI-IOT vLLM on older Orin.

6. 6jetson-packageverified and signedpassport

   skillnvidia/skillsrecord

   • jetson-package

     Pick Jetson-compatible containers, vLLM runtime images, and Jetson AI Lab PyPI indexes; maps Orin SM 8.7 vs Thor SM 11.0 and JetPack-specific package choices.

7. 7jetson-speculative-decodingverified and signedpassport

   skillnvidia/skillsrecord

   • jetson-speculative-decoding

     Add EAGLE-3 or draft-model speculative decoding to a Jetson vLLM server when TPOT is the bottleneck.

8. 8jetson-llm-mcpfound in code, not yet checkedno passport yet
   • ask_jetson

     Ask a local model on Jetson AGX Xavier (jetson-xav2, port 11434). The first call after an idle period adds ~30s of model loading. Models: qwen3.6:35b-a3b (default), qwen3.6:27b, qwen3:30b-thinking. Switching model evicts the loaded one — only one fits in memory. qwen3.6 reasons by default and would

9. 9llm-evaluationverified and signedpassport

   skillwshobson/agentsrecord

   • llm-evaluation

     Implement comprehensive evaluation strategies for LLM applications using automated metrics, human feedback, and benchmarking. Use when testing LLM performance, measuring AI application quality, or establishing evaluation frameworks.

10. 10LLM Benchmark MCP Serverfound in code, not yet checkedno passport yet
    • get_pricing

      Vergleicht die Preise von LLMs über OpenRouter.