Open Source
Family of open-weight models with hybrid Mamba-Transformer architecture and 1M-token context window. Designed for specialized AI agents with multimodal support across video, audio, image, and speech.