chifty / AI infrastructure tools
Your AI.
Your infrastructure.
Which models fit your hardware? What does your AI project need? Find out with three free tools for informed technical decisions.
Where do you want to start?
Choose your starting point. Keep the big picture.
GPU server calculator
I have a model. How much GPU memory and RAM do I need?
Calculate hardwareAI model finder
I have hardware. Which models fit my requirements?
Find modelsAI software comparison
I have a use case. Which software can support it?
Compare softwareGood decisions need context.
Three starting points for AI infrastructure
Start with a model you want to run or the hardware you already own. The GPU calculator estimates memory requirements. The model finder filters models against your requirements. The software comparison explains the roles of runtimes, libraries and complementary interfaces.
Transparent, not a black box
The tools use a local, versioned catalog. Searching and calculating need no external model queries. Sources and assumptions remain visible. Memory figures are planning estimates, not speed guarantees. Validate your model, software and workload before purchasing hardware.
For PCs and businesses
Plan text and chat, vision, image generation, transcription and speech synthesis. The tools cover PCs, Apple Silicon and individual servers. Training, fine-tuning, video generation and distributed server clusters are outside the calculation scope.
Frequently asked questions
Do I need an account?
No. All results are freely accessible. Your requirements stay in this browser unless you explicitly share the configuration.
Does sufficient memory guarantee speed?
No. Speed requires a matching measurement. The tools show memory, compatibility and performance separately.
How are planning values derived?
Published hardware guidance takes priority. Otherwise, the calculator derives a conservative value from checkpoint size or parameter count, precision and task-specific working memory. The result links its basis and remains distinct from a measured benchmark.
