Cheapest hosted models

Ranked by the cheapest like-for-like listing we hold for each model — standard tier, at the widest context that model is served at, within one quantisation — one real offer per row with the host named. A model page may show a cheaper row that is quantised or serves less context. Price is the only axis here: a row near the top is cheap, not good. Prices sync daily. Right now the top pick is Llama 3.2 1B Instruct ($0.020/M out). Every row links to full specs, hardware verdicts and provider pricing.