Inference vs. Fine-Tuning
Inference and Fine-Tuning are the two main ways you interact with an AI model, and they place completely different demands on your hardware.
Inference is simply running an already-trained model to get answers—chatting, summarizing documents, writing code, or generating images. Your hardware reads the frozen model weights and produces an output. Nearly all local AI rigs and daily desktop workflows are built around inference, which requires enough VRAM to hold the model plus a modest cushion for your chat context.
Fine-Tuning means taking a pre-trained model and teaching it new habits, specialized jargon, or strict formatting rules using your own custom dataset. Instead of just reading the weights, your GPU has to calculate gradients and actively update them thousands of times. That extra math requires 2x to 4x more VRAM than inference along with heavy sustained compute, which is why most enthusiasts run inference locally on their own desk and rent cloud GPUs only when they need to fine-tune.
Looking for verified hardware rigs?
Check our real-time £/GB VRAM index to find systems optimized for local model inference.