LiveSPX7,620-1.28%NDX29,127-1.41%US10Y4.961%+3.23%BTC76,833-1.21%ETH2,469-2.04%GOLD4,319-1.04%NVDA210.96-8.42%MSFT505.41+1.14%GOOGL349.39+3.23%

AI & Tech

AI glossary

Inference, tokens, context windows, FLOPs, hyperscaler capex and the rest of the AI vocabulary — each explained plainly, with a worked example built from live market figures.

AgentA model given tools and allowed to take steps.BenchmarkA standard test — and an imperfect one.Compute clusterThousands of accelerators wired to act as one machine.Context windowHow much the model can hold in mind at once.Cost per million tokensThe unit price of running AI in production.Data centre powerThe megawatts that have become the real constraint.DistillationTraining a small model to imitate a big one.EmbeddingText turned into numbers that capture meaning.Fine-tuningSpecialising an existing model on your own data.FLOPsA count of the raw arithmetic a model consumes.Foundation modelA general model others build products on.GPU / acceleratorThe chip that makes model maths fast enough to sell.HallucinationFluent output that is simply not true.Hyperscaler capexWhat the cloud giants spend building AI capacity.InferenceRunning a trained model to produce an answer.LatencyHow long the user waits, split into two numbers.Open weightsA model whose parameters you can download and run.Parameter countHow many weights the model carries.RAG (retrieval-augmented generation)Fetching real documents and putting them in the prompt.TokenThe chunk of text a model actually reads and bills for.TrainingTeaching a model by adjusting its weights on data.