Local AI hardware requirements: more RAM vs big GPU

Geeky Gadgets' Julian Horsey reports that running the Qwen 3.8 Flash Next, a 125 billion parameter large language model, at home requires 96 GB of system RAM, with no GPU memory figure specified, highlighting the trade-off between system memory and GPU capacity for local AI hardware.

Julian Horsey at Geeky Gadgets looks at a question that comes up when running large language models at home: buy more system RAM, or buy a bigger GPU. The example is Qwen 3.8 Flash Next, a 125 billion parameter model. It recommends 96 GB of system memory and gives no GPU memory figure at all. …