Tools · BY Alex Wormuth
Fly Language Model
Browser version not available yet
The language model requires the LFM backbone, trained adapter, and local Python/Torch inference, with roughly 10 GB of model storage documented. Its hosted page blocks embedding. A browser weight format and WebGPU inference port are still needed.
Inspect the original source below ↓Video by Alex Wormuth
Tools / Code available
About Fly Language Model
Talk to the fly. A frozen language model gets a small, trained adapter that reads activity from the fly’s connectivity graph.
Model & implementation
Language comes from LFM2.5-1.2B-Instruct. Only the adapter learns; the anatomical graph stays fixed. The associated study does not establish an advantage from fly anatomy.
Inspect the source without leaving Flymaxxing
Pinned source revision · loaded from the original public repository
Choose a file to read its source.