LLM VRAM calculator
Parameters × quantization → VRAM, with a live match against consumer and datacenter GPUs.
Static tools for AI builders
AI calculators, with the math on the table.
Every number here is computed in your browser from disclosed formulas and vendor-published specs — no APIs, no accounts, no black boxes. Just the conversions AI practitioners actually need.
6 calculators, one data source
GPU specs shared across every tool — tables can never disagree.
Browse by task
What are you planning?
Every tool runs entirely client-side
Tokens per second
Decode speed ≈ memory bandwidth × 0.6 ÷ model size. Compare every GPU.
Training time
6 × params × tokens ÷ achieved FLOPS → GPU-hours for your pretrain or fine-tune.
API cost calculator
Input × price + output × price — one workload, every major provider compared.
Multi-GPU splitter
Total VRAM ÷ card memory → how many GPUs your rig needs.
GPU rental cost
Hourly × cards × hours — the same job priced across RunPod, Lambda, AWS, and Azure.