TensorRT Edge-LLM 在 Jetson AGX Thor 上完成 MLPerf 边缘 Agent 基准，快 6.4 倍
原标题：TensorRT Edge-LLM Completes the MLPerf Edge Agentic Benchmark 6.4x Faster on Jetson AGX Thor
AI 摘要
NVIDIA 在 MLPerf Inference v6.1 Edge Agentic 基准测试中，用 TensorRT Edge-LLM 在单台 Jetson AGX Thor 上运行 Qwen3.6-27B，以 52.33 tokens/秒完成 1,007 轮 agent 工作负载，耗时 24 分 36 秒，比 llama.cpp 参考提交快 6.4 倍。该结果依赖 NVFP4 量化、树状多 token 预测（MTP）和 KV 缓存复用，BFCL 准确率为 87.94%。
正文节选
AI agents are moving from cloud data centers to vehicles, robots, and other edge devices. Unlike a chatbot that answers a single prompt, an agent works through a sequence of steps. It selects tools, evaluates their results, and continues reasoning within an increasingly long conversation. This workflow places new demands on edge inference. The model must generate tokens quickly, process long shared histories efficiently, and produce valid tool calls within a limited power and memory envelope. In