reddit · SideProject · 09-29 06:50 51 分 ← 返回列表

I built a tool for finding local LLMs that fit your hardware

I've been wanting to experiment with different local LLMs, but I kept running into the same problem: figuring out which models were good for what I wanted to do, and which ones I could actually run on the hardware I had. So I started collecting information about models, memory requirements, context limits and hardware, and eventually turned it into Local LLM Finder. You choose your hardware and what you want to use the model for, and it gives you a shortlist of models that should fit. You

五维评分

付费意愿

需求真实度

个人可落地

竞争空间

新鲜度

AI 分析

痛点：本地 LLM 爱好者难以判断哪些模型能在自己硬件上运行、适合什么用途。

目标用户：拥有消费级 GPU/Apple Silicon 的本地 LLM 玩家、AI 开发者、需要私有化部署的小团队。

证据：figuring out which models were good for what I wanted to do, and which ones I could actually run on the hardware I had.

变现：硬件/云 GPU 联盟导购、高级筛选与基准数据订阅、API 嵌到模型库、广告。

MVP：两周内做硬件选择器+模型数据库：输入 GPU/内存/用途，输出可运行模型列表、量化建议、预计 tokens/s，并支持模型对比。

竞品：LocalLLM Finder、Ollama model library、LM Studio、GPT4All、Hugging Face model cards、llmfit

摘要：发帖人因难以判断本地LLM与硬件匹配，用AI辅助做了Local LLM Finder，按硬件和用途推荐可运行模型，并计划改进估算数据。帖子未提及付费或收入，主要是求反馈，变现路径偏联盟导购/订阅。