ITG NowHome › AI glossary › Inference
What is inference in AI?
Short answer
Inference is when a trained AI model is used to answer a question or make something, as opposed to learning.
- Category Models
- Updated 20 Oct 2026
Inference, explained simply
Every time you send a message to a chatbot, the model runs once to answer. That run is inference.
Training happens once, in big data centers. Inference happens every time anyone uses the model.
That is why running popular AI costs real money. It adds up across millions of chats.
On ITG Now, our free tool guides explain how to run inference on your own computer with local tools.
Related terms
TrainingLatencyTokens per secondGPULocal AIAI model
Where you'll see it on ITG Now
- AI newsDeepSeek open-sources DeepInfer, the engine behind its servers
- AI newsChip startup Tenstorrent raises $1.5B at $6.2B value
- AI newsReflection Beam: open coding model arrives on AWS
- AI newsDeepSeek says free model update tops a coding test
- AI newsOVHcloud plans its own AI model in early 2027
Questions people ask about Inference
Inference vs training: what is the difference?
Training teaches the model. Inference uses what it learned to give you an answer.
Can I do inference on my own laptop?
Yes, for small and medium models. Bigger ones need more memory and a good graphics card.