MAI
mai_gen_b1
and mai_gen_p1 (coming soon)
are powerful LLMs trained specifically on crypto / defi datasets. They also include the ability to obtain live data and relay it to the user conversationally.
They each utilize a series of "sub-models" which are trained and optimized on certain categories, some of which can be seen below:
Token Activity (ex. price fluctuations, growth)
Token Information (ex. safety)
Trader Activity
Trader Information (ex. pnl changes, inflow/outflows)
Solidity / Rust Contract Information

Scan
The Scan page offers insight into some of the live data we are using to train + fine-tune MAI.
This feed is split into to sections, first the console output, and the formatted. The console output gives more technical insights, while formatted is adjusted to be read easily users.
More and more datasets will be integrated weekly, causing MAI to grow more powerfully each training session - until Mainet is live. In which the new protocol will take over.
To view what we currently use to train our models, visit Training Mai.
To start chatting with MAI, visit our web-app or telegram bot
Last updated