SubQ Snap launches enterprise AI model with 384K token context

SubQ Snap: SubQ Snap is a multimodal decision model developed by Subquadratic for fast, long-context reasoning in enterprise agentic workflows. It builds on the company's Subquadratic Sparse Attention architecture to maintain constant compute per token regardless of context size. The model was launched today as the company's first dedicated enterprise decision model following its earlier SubQ LLM release. Alex Whedon: Alex Whedon is the co-founder and CEO of Subquadratic, an AI company focused on developing subquadratic foundation models. He announced the launch of SubQ Snap and leads the team's efforts to advance efficient scaling in large language models for practical enterprise applications. Industry Category: Long-context decision models represent an emerging category aimed at delivering efficient, accurate outputs for enterprise benchmarks spanning tool use, security, legal, finance, and code tasks. Model Architecture: Subquadratic's approach emphasizes architectural innovations that address compute and memory scaling challenges in foundation models, enabling more practical long-context processing for real-world use cases. Enterprise Workflows: Decision models are increasingly positioned to handle structured tasks in agentic systems where full context from codebases, documents, or histories improves reliability without relying solely on larger general-purpose LLMs.