CostPerPrompt

CostPerPrompt is a pricing tracker and calculator built for developers and teams who need to understand what AI APIs will actually cost before the invoice arrives. It covers 232 models from providers including OpenAI, Anthropic, Google, DeepSeek, and Qwen, with live pricing data that updates daily. The core pitch is simple: most cost estimates run two to three times too high or too low because they ignore prompt caching and batch processing discounts. This tool factors those in.

The problem it addresses is common once you start building on LLM APIs at scale. You pick a model, estimate token counts, multiply by the published rate, and end up surprised when the bill doesn't match. Prompt caching can cut input costs by up to 90 percent on repeated context, and batch endpoints knock off around 50 percent for non-realtime workloads. If your estimate ignores those, it's fiction. CostPerPrompt works those discounts into the math so the number you see is closer to the number you pay.

The calculators are organized by workload type. There's a general API cost calculator where you plug in requests per day, average input tokens, and output tokens, then compare across models. Beyond that, dedicated calculators exist for chatbots, RAG pipelines, autonomous agents, voice AI applications, GPU rental, and image generation APIs. Each calculator surfaces the parameters that actually move the needle for that specific use case, which makes it faster to explore the cost landscape without building spreadsheets from scratch.

For anyone building production AI features, this is the kind of reference you'd otherwise compile internally or piece together from provider docs. It pulls pricing from official sources, shows you when it last refreshed, and lets you report corrections directly if something drifts. The value is in the aggregation and the specificity of the calculators, not in proprietary data.

Who it fits best is product managers scoping an AI feature, engineers comparing model families during vendor selection, or finance teams trying to forecast spend before a feature ships. If you're past hobby scale and negotiating budgets, having a single place to sanity-check numbers helps conversations move faster.

The site is free to use and supported by advertising and affiliate links. It's independent and not tied to any AI provider, so the comparisons aren't weighted toward a particular vendor. There's no account required to access the calculators or the model comparison tables. You just open the page, enter your workload assumptions, and get numbers.

One thing to note is that pricing changes fast in this space, sometimes weekly. The tool shows last-updated timestamps and invites corrections, but if you're making a major budget decision, you'll still want to double-check against the provider's official page before committing. That said, as a starting point for narrowing down which models to evaluate, or for catching savings opportunities like caching you might have overlooked, it's a useful resource to have bookmarked.