# What is Inference? Definition & Technical Mechanics

> **Source:** [SPIDITS AI AI Glossary](https://spidits.com/ai-glossary)  
> **Category:** Model Operations  
> **Canonical Citation:** [https://spidits.com/ai-glossary/inference](https://spidits.com/ai-glossary/inference)

## Definition
Inference is the process of using a trained AI model to make predictions or generate text based on new inputs. During inference, data flows forward through the neural network to produce an output, without modifying the model's weights.

## Architectural Mechanics & Deep Dive
Inference is the phase where a trained machine learning model runs in production to process new, unseen inputs and generate predictions or content. Unlike the computationally expensive training phase, inference does not update model weights; it simply passes the input through the network to calculate the output. Optimizing inference (e.g., using quantization or cache systems) is essential for reducing serving latency and hosting costs.

## Industry Applications & Real-World Use Cases
Real-time chat responses, API query execution, and mobile-device AI features

## Frequently Asked Questions

### How does inference differ from training?
Training calculates errors and updates the model's weights. Inference just uses the already-calculated weights to process new data.

### Why is inference speed important?
Low latency inference is critical for interactive user experiences (e.g. conversational voice agents).

## Related Terminology
- [models](https://spidits.com/ai-glossary/models)
- [llm](https://spidits.com/ai-glossary/llm)
- [loop-engineering](https://spidits.com/ai-glossary/loop-engineering)

---
*From the SPIDITS AI Knowledge Base (336+ terms indexed). Explore full technical definitions: [https://spidits.com/ai-glossary](https://spidits.com/ai-glossary)*