Today, we’re launching Assistive Chat, a multimodal assistant with memory. Everyone can start using Assistive Chat today at assistive.chat/chat.

Multimodality

Assistive Chat can create and reason across various modalities, including text, images, videos, and audio.

This means you can simply type to create images and videos, or turn an image into a video.

Later this year we’ll be introducing audio capabilities, allowing you to create audio from text, images, and videos.

And, Assistive Chat can see and reason with images and videos. For example, you could turn a video into code, and Code Preview allows you to try code without leaving the conversation.

Memory

Chat can remember context across conversations. This means it can recall facts and context about your life, work, and environment. With Memory, Chat will be able to provide even more personalized and helpful responses.

And, later this year we'll be introducing Proactivity, which allows Chat to message you first with helpful ideas and suggestions.

Analysis

Chat can analyze data and offer actionable insights with Analysis. And, it can perform precise calculations allowing it to solve math problems with better accuracy.

Analysis uses a sandboxed Python interpreter that can handle uploads and downloads. Unlike ChatGPT's Code Interpreter, sessions do not time out, meaning you can pick up where you left off anytime.

Retrieval

You can upload and chat with your PDF documents. Ask questions, get summaries, find information, and more.

Through retrieval, your assistant can work with documents that traditionally exceed the limited context length of large language models.

Browsing

To provide more relevant and up-to-date answers, Assistive Chat can search the web and browse webpages. This allows Chat to stay updated with knowledge not available in its model's training data.

And, you can also see traditional search results below the assistant's response to explore further.

Next steps

We'll be working to make Assistive Chat even more helpful, and later this year we'll be making audio capabilities and Proactivity available. We'll also be improving Memory, and introducing more ways to provide personal context to your assistant.

And, we'll also be introducing some features such as Analysis and Retrieval to the Assistive API, so developers can include these technologies in their own applications.