Aquila is a new family of AI Models designed by me designed to be intelligent and very easy to run. Aquila began as a project in early september and the first time it was talked to was October 13th.
Aquila models run in a custom inference engine called UAL (Ultimate Aquila Loader) which is built and compiled entirely in Fortran. The model files are encrypted in a custom format that only me and my advisor know.
Aquila M1 (Model 1) completed training on October 9th after beginning around a week prior and Aquila M2 is currently in training.
The normal Aquila models are currently 9b but a much larger model family called Aquila Heavy (50+b) is going to be trained by my advisor using data center technology in the future.
A major future goal is Agentic Aquila which will have full unrestricted access to the internet and be fully autonomous.
Aquila will be integrated with my Tuxron game engine and will be able to run on cpus and gpus efficiently, currently only runs on CUDA tho.
Aquila is not available to the public.