# Time to First Token (TTFT)

> How long you wait before the first word of an AI answer shows up.

**Field:** Artificial Intelligence  
**Category:** LLMs  
**Also known as:** TTFT, First Token Latency

## Think of it like

It is like the time from when you ring a bell to the first hello.

## Definition

Time to first token is how long a language model takes to send the first piece of its answer. The clock starts when you send the request and stops when the first token arrives. In that gap, the request may wait in line, the model reads your whole prompt, and it may think first. After that, the rest streams in at a steady rate. A long prompt or a thinking step makes the gap longer.

## Example

Jack's chat app shows the first word after 0.4 seconds, then streams the rest. Users feel it is fast, even though the full answer takes 6 seconds.

## Why it matters

It is the wait people feel most. A short gap makes an app feel fast, even when the full answer is long.

## Common mistake

Measuring only the total time. Two apps can finish together, while one feels far slower to start.

## Three principles

- **The first gap.** Send to first token.
- **Prompt size counts.** Long input means a longer wait.
- **Then it streams.** Speed after that is per token.

## Related terms

- [LLM Streaming](https://learn.benibix.com/term/ai/streaming-response)
- [Token (AI)](https://learn.benibix.com/term/ai/token)
- [KV Cache](https://learn.benibix.com/term/ai/kv-cache)
- [Prompt Caching](https://learn.benibix.com/term/ai/prompt-caching)
- [Inference (AI)](https://learn.benibix.com/term/ai/inference)
- [Extended Thinking](https://learn.benibix.com/term/ai/extended-thinking)

## Taught in

- [LLM Speed and Streaming](https://learn.benibix.com/learn/ai-llms/llm-speed-and-streaming): LLMs, 13/15

![A timeline from left to right. You send, then a highlighted stretch labelled time to first token, where the request waits in line, the prompt is read and the model may think. Then the first word arrives, and the rest streams in at a rate set by tokens per second, until done. A note says a long prompt or a thinking step makes the first gap longer.](https://learn.benibix.com/diagrams/ai2-ttft.svg)

---

Source: https://learn.benibix.com/term/ai/time-to-first-token
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-18
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
