Browse by category
LLM Inference Cost & Serving
What it costs to actually run a language model in production: per-token API pricing split across input and output, cost per user…
Model Memory & GPU Sizing
Whether a model fits, and on what: weight memory at each precision, KV-cache growth with context length and batch size, the…
Training & Fine-Tuning Compute
The arithmetic behind a training run: parameter counts from architecture, the FLOPs a run consumes, how long it takes on given…
RAG, Embeddings & Vector Search
Sizing a retrieval-augmented generation system: how a corpus divides into chunks, what embedding it costs, how much a vector…
Model Evaluation & Metrics
Turning raw counts into the numbers a model is judged on: precision, recall and F1 from a confusion matrix, and perplexity from…
Most used in ai, llm & machine learning engineering
- LLM API Token Cost CalculatorPer-million-token billing conventionLarge-language-model APIs bill per million tokens, with output tokens costing several times more than input tokens and cached…
- LLM Cost Per User Per Month CalculatorToken-metered unit economicsThis calculator turns per-token API pricing into the number that decides whether an AI feature can be priced profitably: what…
- Self-Hosted LLM vs API Break-Even CalculatorFixed-versus-variable cost break-even analysisSelf-hosting an open-weights model swaps a variable per-token bill for a fixed hourly one, so it wins above a volume and loses…
- Prompt Caching Savings CalculatorPrefix-cache token pricing modelCaching a long, unchanging prompt prefix is not free: the first request pays a premium to write the cache, and every hit after…
- LLM Throughput & Latency Calculator (Tokens Per Second)Memory-bandwidth-bound decode modelToken generation on a transformer is bound by memory bandwidth, not by arithmetic: every token generated requires reading the…
- LLM GPU VRAM Requirement CalculatorThis calculator adds up the four things that occupy GPU memory when you run a language model: the weights, the KV cache, the…
- KV Cache Size CalculatorTransformer key-value cache memory modelEvery token a transformer has already seen leaves a key and a value vector in memory for each attention layer, and that cache…
- Model Quantisation Memory Savings CalculatorWeight-precision memory accountingA model's weight memory is just parameters times bytes per parameter, so quantisation buys memory in exact proportion to the…
- GPU Hours Cost CalculatorGPU-hour cost accountingA GPU-hour is the unit every cloud accelerator bill is written in: one GPU held for one hour. This calculator turns an hourly…
- Transformer Parameter Count CalculatorGive this calculator the six numbers from a model config — hidden size, layer count, feed-forward width, attention heads…
- LLM Training Compute (FLOPs) CalculatorC = 6ND scaling estimate (Kaplan et al. 2020; Hoffmann et al. 2022)This calculator turns a parameter count and a token budget into the three numbers that decide whether a pre-training run is…
- Fine-Tuning vs Long-Prompt Break-Even CalculatorFixed-versus-variable cost break-even analysisFine-tuning trades a one-off training bill, and usually a higher per-token serving rate, against never sending your few-shot…