# Common Architectures

Some of the most common applications that can be built on Stack AI rely on different components of our product. Below are some of the elements to look into for solving each task:

| LLM Task                                      | Recommended LLM(s)                      | Useful Components & Architecture                                                              |
| --------------------------------------------- | --------------------------------------- | --------------------------------------------------------------------------------------------- |
| Ask questions to several documents.           | `GPT-3.5-turbo or Claude-v1`            | Offline Data Loaders: `Websites`, `Data + Search` `URLs + Search`                             |
| Browse the internet to perform research       | `Claude-v1-instant` and `GPT-3.5-turbo` | Data Loaders & Vector Databases: Input, LLM, Google Search, VectorDB, LLM, Output             |
| Aggregate data from a database or table       | `GPT-3.5-turbo-16k` or `GPT-4-32k`      | Data Loaders & Document Readers: Input, Data Loader (Airtable, CSV, etc...), Doc Q\&A, Output |
| Perform operations on a table or database     | `GPT-4`                                 | Plugins: Input, Table Analyzer, LLM, Output                                                   |
| Transcribe a document into a different format | `GPT-3.5-turbo-16k`                     | Document Readers: Input, Data Loader, Transcriber, Output                                     |

Explore all the different components and build over these architectures!