> For the complete documentation index, see [llms.txt](https://knowly-ai.gitbook.io/langtorch/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://knowly-ai.gitbook.io/langtorch/capability/capability-unit.md).

# Capability Unit

When you are sending queries in ChatGPT, you are basically sending queries to a large language model with a pre-defined prompt. In langtorch, we call it "**capability unit**" or in real life, it is roughly equal to "**a** **task**", i.e. performing a Q\&A task.

<img src="/files/clfuwDUHcuHt12vjni9N" alt="" class="gitbook-drawing">

However, **LLM is NOT ALWAYS** [***all you need***](https://arxiv.org/abs/1706.03762) [😂](https://emojipedia.org/face-with-tears-of-joy/) and sometimes, you don't even bother using a LLM. For example, involving LLM when **calculating (1287652 \* 2372837) / 7.8765 + 9876** is not a good idea even with GPT-4 and the famous magic "[Let's think step by step](https://arxiv.org/pdf/2205.11916.pdf)".

<figure><img src="/files/4tDZh3EulsUmZCoTHlWy" alt=""><figcaption></figcaption></figure>

The true label is 387911938870 if you are curious.

For cases like this, a calculator function is preferable.&#x20;

We also have built-in support for functions.

To make it simple with our langtorch jargon,&#x20;

**The capability unit is ONE OF:**

1. **Prompt Template + Model**&#x20;
2. **Functions**