Genus
A native large-model inference engine written in pure C. It loads Mixture-of-Experts (MoE) expert weights by streaming them from an SSD, which allows devices with limited memory to run models with very large parameter counts.
Timeline (last 90 days)
Genus is an inference engine written in pure C, and the author is Janas.
The core mechanism of Genus is to keep model weights on the SSD without loading them into memory, and each token only reads the activated expert weights from the SSD.
On the same 32GB laptop, running the model in the traditional way with the CPU maxed out gives a speed of about 2.8 tokens/s.
The author Janas explicitly stated that Genus is just an experiment, not yet a product.
Genus most likely lacks complete documentation, error handling, multi-platform support, and a long-term maintenance commitment.
All Facts (5)
Genus is an inference engine written in pure C, and the author is Janas.
50%UnverifiedThe core mechanism of Genus is to keep model weights on the SSD without loading them into memory, and each token only reads the activated expert weights from the SSD.
50%UnverifiedOn the same 32GB laptop, running the model in the traditional way with the CPU maxed out gives a speed of about 2.8 tokens/s.
50%UnverifiedThe author Janas explicitly stated that Genus is just an experiment, not yet a product.
50%UnverifiedGenus most likely lacks complete documentation, error handling, multi-platform support, and a long-term maintenance commitment.
50%