AI Rate CardTools: pick a model by cost
Find the cheapest offer for your monthly workload, count the tokens of a prompt, convert price units.
Cost picker
How many tokens a month you send to the model (input) and get back (output), in millions. For example, a chatbot with 1,000 conversations a day is about 10 M input and 2 M output. Don't know your numbers? Count them below in the token estimator.
Model: granite-7b-base ×
| Model | Provider | Your cost/mo, $ | Input, $ | Output, $ | Context |
|---|---|---|---|---|---|
| granite-7b-base | featherless | $0.56 | $0.40 | $0.80 | 4K |
Images, video and voice cost calculator
How much a job costs at every model that sells it by the image, the second of video, the character of speech or the minute of transcription - today's prices, the cheapest offer of each model.
Budget calculator: what $N a month buys
Type your monthly budget and a typical request: how many requests a month (and a day) it buys on each model at its cheapest seller, with the model's reasoning class.
RAG indexing cost calculator
What document search for an AI assistant (RAG) costs in embeddings and reranking: your documents embedded once, the changed share every month, the queries, and the passages reranked - at the cheapest seller of every model.
Coding agent cost calculator
What a coding agent costs a month on each model: tasks a day, model calls per task, the context each call sends, its answer and reasoning, and the share read from the prompt cache - at the cheapest real-time seller of each model.
Subscription or API: which is cheaper
A chat subscription's monthly price against the same use paid by the token through the API: at the cheapest seller of the model and at the model maker's own API, and how many messages a day make the subscription pay off.
Prompt cache and batch savings calculator
Your monthly tokens at every seller of one model: the plain price, the price when part of the prompt is read from the cache, and batch / flex offers if the job can wait.
Model classes S/A/B/C/D
How strong a base model is in one direction, as a letter: S is the top, D the bottom. Each letter stands on at least two independent public sources; one source gives a score without a letter.
Token estimator
Don't know your workload in tokens? Paste an example of your prompt, say how many such requests you make a month and how long an answer is: you get the monthly workload. The button puts it into the picker above and shows the offers.
Price units
Providers quote per token, per 1K or per 1M tokens: the same price in all three.
AI picker
Describe the task in your own words - what the model must do, how fast, the budget, the region - and get a short list of models and providers with the reasons.
Coming soon.