SpectralFoundation Model Series / Spectral 1
Aethr Intelligence’s flagship foundation model series developed on Project OMN architecture.
Spectral (with initial release designated Spectral 1) is Aethr Intelligence’s proprietary foundation model series. It operationalizes our Project OMN architecture and geometric training research toward models engineered for frontier reasoning, engineering domain specialization, and verified edge silicon execution.
Our priorities are efficient geometric training, rigorous empirical validation, and safe superintelligence design. The historical OMN-1 prototype informs this work, establishing structural foundations for the forthcoming Spectral release sequence.
Spectral 1
An OMN foundation
The model series is being developed around our architecture and training research. OMN-1 is a dense decoder-only Transformer, not a demonstrated replacement for the Transformer paradigm.
Capability in context
We are interested in models that can be adapted to technical domains and evaluated on well-defined tasks. A model name alone does not establish domain expertise.
Evidence before claims
Training stability, context use, memorization and instruction-following require separate tests. Historical OMN results should not be read as validated Spectral 1 release capabilities.
Details when verified
Model sizes, context limits, licenses, access arrangements and benchmark results will be published with the relevant releases. We are not announcing availability or performance guarantees here.
Discuss model research with Aethr.
We welcome relevant research and domain discussions. Model access, checkpoint availability and release terms have not been announced.