#本地模型

5 tutorials, projects and news items tagged 本地模型.

The Uncensored Gemma 4 12B Is Taking Off on 12 GB GPUsPick

Gemma 4 12B Heretic is a community model that uses abliteration to strip out the refusal behavior of Google's Gemma 4 12B, and it runs on a 12 GB graphics card in GGUF form. It has been downloaded more than 100,000 times in the past 30 days, but Google did not release it and the repo publishes no benchmark data. This guide covers what it is, which files fit which hardware, which claims hold up, the risks, and a prompt that lets your AI agent help you pick a local model.

News·Beginner·4 min·Nov 18, 2026gemma 4gemma 4 12b本地模型gguf
Sell Your Idle Local AI Box While Used Prices Hold Up

A 128GB DGX Spark handles image and video generation and small models, but falls short on mainstream big models. It is a personal view, not an official claim, that 128GB-class memory will soon be common, so selling idle hardware while resale prices hold makes sense.

Experience·Beginner·1 min·Membersdgx spark本地模型统一内存硬件
That cheap used workstation may cost more once you add the missing parts

Cheap used workstations are often barebones with no graphics card, memory, or drive, and the OS install can get stuck too. Make a parts list before you add up the total. Includes a prompt that has your AI agent do the math.

Experience·Beginner·1 min·Members二手工作站准系统本地模型硬件
Hunting for a cheap used GPU? Make sure it can actually run AI first

Used graphics cards and ex-mining cards look like bargains, but the model, warranty, and cooling matter more than the price. Only a few mining cards have been unlocked, and changing the memory can make heat hard to control. Includes a prompt for checking a card before you buy.

Experience·Beginner·1 min·Members二手显卡矿卡显存本地模型
Before you buy big RAM for a local LLM, run the numbers first

Lots of RAM only means a model fits, not that it runs well. Before you buy hardware, check speed, how the RAM is set up, and model quality, then compare a year of cost against a subscription. Includes a prompt to run the numbers.

Experience·Beginner·1 min·Members本地模型local llm内存硬件