Models
A family of frontier models built on the same silicon foundation, scaled from edge to supercomputer.
Silkon 1T
AvailableOur flagship model with 1T parameters, trained on diverse high-quality data.
FlagshipMultimodal
Params
1T
Context
128K
Training
10T tokens
Silkon 7B
Coming SoonCompact yet powerful model optimized for edge deployment and high-throughput inference.
EfficientOpen Weights
Params
7B
Context
32K
Training
2T tokens
Silkon 400M
Coming SoonDesigned for mobile and embedded applications without sacrificing reasoning quality.
On-DeviceEdge
Params
400M
Context
8K
Training
500B tokens