ChainCatcher 消息，Vitalik Buterin 发文分享其截至 2026 年 4 月的本地化、私有化 LLM 部署方案，核心目标是将隐私、安全与自主可控作为前提，尽量减少远程模型及外部服务接触个人数据的机会，并通过本地推理、文件本地存储及沙箱隔离等方式降低数据泄露、模型越狱及恶意内容利用风险。
在硬件方面，其测试了搭载 NVIDIA 5090 GPU 的笔记本、AMD Ryzen AI Max Pro 128 GB 统一内存设备及 DGX Spark 等方案，并使用 Qwen3.5 35B 与 122B 模型进行本地推理。
其中，5090 笔记本在 35B 模型下可达约 90 tokens/s，AMD 方案约 51 tokens/s，DGX Spark 约 60 tokens/s。Vitalik 表示，其更倾向于基于高性能笔记本构建本地 AI 环境，同时使用 llama-server、llama-swap 及 NixOS 等工具搭建整体工作流。
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.