Compute Prices

About Compute Prices

Compute Prices is a price index for AI infrastructure, tracking what GPUs and LLM inference actually cost across the market and publishing the answer as both a browsable site and a JSON API. The scope is the useful part: roughly eighty providers and several hundred thousand price points refreshed daily, covering sixty-seven GPU providers across seventy-four GPU models, and twenty-six LLM inference providers across more than three hundred models — Lambda, Modal, Runpod, Vast.ai, Together AI, Nebius, Paperspace, Salad, Scaleway, Vultr, OpenRouter, Deep Infra, Fireworks and the hyperscalers among them. For anyone sizing a training run or picking an inference vendor, the alternative is scraping a few dozen pricing pages with a few dozen different layouts, which is exactly the work this removes. Beyond the raw table it computes the derived views people actually want: biggest price drops this week, lowest current price in each category, and per-tier week-over-week movers with daily trend history. The API is positioned explicitly for AI agents — one stable response shape across every provider, so a cost-aware agent can query current GPU or token pricing as a tool call rather than parsing HTML. A free key takes about thirty seconds and allows 750 requests a day with weekly movers and seven-day trends; a Pro tier extends history to ninety days. The site is candid that prices are collected automatically from public sources and provider APIs and should be confirmed with the provider before purchase.

Key Features

Tags