Foundational language models
Compact models for reasoning, generation, structured outputs, and tool use.
Coming soon
Models
We develop compact models and inference systems for workloads where latency, cost, privacy, and deployment constraints matter. Selected models, evaluations, and technical results will be published as they mature.
Model areas in development
Retrieval systems
Search, retrieval, reranking, and generation over private and domain-specific knowledge.
Document understanding
Models for extracting text, structure and entities from documents, forms etc.
Physical AI & robotics
Perception, planning, and decision models for machines operating in physical environments.
Speech
Models for transcription, synthesis, speaker understanding, and real-time voice interaction.
Have a workload existing models don’t fit?
Talk to our research team →