CASE EVIDENCE / A RECORD
Neural Magic / Red Hat AI 公开的智能体工作流案例，来源为 huggingface，复核于 2026-06-27T08:07:03Z。
原始记录：Neural Magic / Red Hat AI created an FP8 vLLM-ready Qwen2 72B Instruct artifact
Chinese Brief
Neural Magic / Red Hat AI 公开的智能体工作流案例，来源为 huggingface，复核于 2026-06-27T08:07:03Z。 Model Atlas 将它标记为 A 类证据，因为它同时具备具体使用者、具体任务、公开原始证据和可访问产物。 Model Atlas 不把 benchmark、教程、发布说明或集合页包装成真实案例。
任务
这是一个围绕智能体流程编排的真实任务，公开材料可以回溯到具体使用者和具体产物。 原始资料写作：Neural Magic quantized Qwen2 72B Instruct weights and activations to FP8 and published a Red Hat AI Hugging Face artifact intended for assistant-like chat deployment with vLLM.
公开材料提供原始证据链接和可访问产物，可用于核验任务结果、项目形态和模型绑定关系。 原始资料写作：The public model card states the optimized artifact is ready for inference with vLLM >= 0.5.0 and reduces disk size and GPU memory requirements by approximately 50% compared with the original precision.
Qwen2 72B 在该案例中承担智能体流程编排相关的生成、分析、编排或实现角色。 原始资料写作：Qwen2 72B Instruct provides the base instruction-following model; the case demonstrates its adaptation into a lower-memory FP8 deployment artifact.
当前判断基于公开材料；若产物下线、仓库变更或模型参与比例仅来自作者自述，需要在引用前重新复核。 原始资料写作：This is a concrete public derivative/deployment artifact, not an end-customer story; evidence explicitly says it is a quantized version of Qwen2-72B-Instruct.