What it costs
Local AI is a fixed cost with a low marginal cost. These numbers use your machine, your usage, and current cloud prices.
Machine
Local
/mo
/mo
Total spend over 12 months
The dot is the month the local rig catches up.
At 12 months
Local uses the speed estimate for the best model, one device, 8k context. Cloud prices come from models.dev for GPT-5.5 and DeepSeek V4 Pro, with four input tokens per output token. Hardware is amortized over its life; electricity is the only running cost. Speeds are estimates, not benchmarks.