Free AI cost tools
Free, instant, no signup. Everything here runs in your browser — nothing you type is uploaded, and no API key is involved. Rates are read live from the Kunavo catalog, and every model is shown next to its provider's own list price so the numbers can be checked rather than trusted.
Price any workload across every Claude, GPT and Gemini model side by side, each next to its provider's list price.
OpenClaude rates with the cache maths the others leave out — cache hits at a tenth of input and cache writes at 1.25×, each at the model's own discount.
OpenPer-clip and per-second video models priced as the different things they are — resolution, clip length and volume, each next to the provider's list price.
OpenEvery image model on the key, priced per image and per tier — one is flat across resolutions and the rest are not, so the cheapest model changes with the size you ship.
OpenGPT rates including cached input at a tenth of the input rate, and what reasoning tokens do to an output-heavy bill.
OpenWhy there are three calculators and not one
Because the billing rules differ, not the branding. The cross-vendor calculator is the right tool when the question is “which model is cheapest for this workload”. The two vendor calculators exist because prompt caching — the term that dominates any agent or editor bill — works differently on each side: a cache hit costs a tenth of the input rate on both, but every Claude model bills its cache writes at 1.25× input, while on the OpenAI side only GPT-5.6 and GPT-6 Astra have a cache-write charge to reason about.
The media calculator is the sharpest case of the same rule. Video models do not even agree with each other: some bill per clip, where resolution moves the price and length does not, and some bill per second, where length is the dominant term. One form covering both would be wrong for half the catalog, so the video calculator reads the billing unit per model and asks for a duration only where duration is billed.
Images have no duration lever at all — every image model bills per output — so the image calculator turns on resolution instead, and on the fact that one model is flat across tiers while the rest are not. Change the tier and the cheapest model changes, which is the sort of answer a table cannot give.
Kunavo is an independent, OpenAI-compatible AI API gateway: one API key and one pay-as-you-go balance reach Claude, GPT and image, video and audio models, with no per-provider account required. That is also why every calculator here can show a second column: the provider's published list price next to what the same workload costs through Kunavo.