Up to 7.2× the speed of standard llama.cpp Runs 100% on your PC
Free for personal use
Download free
How-to guides/Guide 04
04

Chat basics

5 steps · about 4 minutes

Chatting with 5090Ai works like any AI assistant, only much faster, and everything stays on your PC.

  1. Start a new chat

    Click New chat (or press Ctrl + N). You can click one of the suggestions to try it out.

    5090Ai
    A new chat with suggestions
  2. Type your message and send it

    Type in the message box at the bottom and press Enter. For a new line, press Shift + Enter.

    The answer appears as it's written. The speed meter at the top right shows how fast, next to standard llama.cpp's speed for the same model. To stop early, click the stop button or press Esc.

    5090Ai
    An answer being written, with the live speed meter at the top
  3. Read and copy the answer

    Answers are formatted, with tables, maths and code. Every code block has a Copy button.

    5090Ai
    A finished answer with code and an explanation
  4. Check the statistics

    Under each answer: its speed, how many times faster it was than llama.cpp, the number of tokens, how long the first word took, and the model used.

    Answer statistics: 772 tok/s, 4.2× llama.cpp, 621 tokens

    Hover over a message to copy it, Regenerate the last answer, or Edit and resend your last message.

  5. Give the chat instructions (optional)

    Click the sliders icon at the top right to set Chat instructions: a short note on how the assistant should answer in this chat. Set a default for every chat in Settings.

    Chat instructions dialog