AI engineering calculators
14 free calculators, no login — type a number, see what it means.
Prompt cache savings
Estimate what caching a long, repeated prompt prefix saves each month.
Response latency
Work out how long a user waits for a streamed answer.
Agent economics
Compare an AI agent with human handling, including the cost of failed attempts.
Retries (pass@k)
See how much retrying or sampling several times lifts the chance of a good result.
Eval confidence
A model passed 45 of 50 test cases. How sure can you be of the real pass rate?
AI coding ROI
Test a "saves 3 hours a week" claim against what teams realistically keep.
Review load
If AI helps people open more pull requests, who reviews them?
GPU capacity
Estimate how many GPUs a self-hosted model needs for your traffic.
Context budget
Check that instructions, retrieved documents and history fit with room for the answer.
Words to tokens
Quick converter for sizing prompts and bills. English text averages about 4 characters or
Retrieval quality
Check whether a search or RAG system finds the right documents.
Hallucination and agent success
Turn evaluation counts into two rates, with a confidence range for each.
Vector index size
Estimate the memory or disk needed for an embedding index.
Judge agreement (Cohen kappa)
Check how well an automated judge agrees with human labels, beyond chance.