The first long-context decision model
for real enterprise work.
SubQ Snap is a multimodal decision model built for fast, long-context decisions in enterprise agentic workflows.
Built for structured decisions in agentic workflows.
Make better decisions with long, multimodal context.
Today's decision models see 32K tokens, a sliver of a real codebase. Snap handles up to 384K, enough for the whole repo, data room or agent history, so every call is grounded in all the context it needs.
Stop paying frontier prices for routine decisions.
Most agent steps are simple calls: pick a tool, open a file, check a result. Snap handles them at a fraction of the cost and leaves only the hard work to your LLM.
Seconds, not minutes.
A decision that takes an LLM minutes takes Snap seconds. Your agents move faster without giving up a single point of accuracy.
The industry leader in short and long context decision-making.
Long-context decisions
JevShort-context decisions
JevEnterprise workflows
Built with
Efficiency on Demand.
What is a long-context decision model?
Decision models are AI models that evaluate supplied context against defined questions and return structured answers. They excel when the deep reasoning or nuanced outputs of a frontier LLM are not required.
| Dimensions | | Language model |
|---|---|---|
| Output | A choice, score, or label | Open-ended text |
| Answer space | Fixed, defined up front | Unbounded |
| Inference | Seconds | Minutes for the same task |
| Parsing needs | None. Structured output your code uses directly | Parse and validate free text |
| Best at | Routing, ranking, classifying, yes/no calls | Reasoning, writing, open-ended work |
| Weak at | Open-ended generation | Fast, repeated, structured calls |
| Place in agent | The decision points: route, filter, score | Planning and generation |
The subquadratic breakthrough.
Today's transformer-based LLMs waste compute by processing every possible relationship between tokens, but only a small fraction of these relationships matter. (This is called “quadratic scaling.”)
SubQ is built differently. Using a proprietary algorithm called Subquadratic Sparse Attention (SSA), it isolates the tokens and relationships that matter, ensuring compute is used efficiently.
Read the ResearchInterested in
enterprise access?
Contact sales.
SubQ 1.1 Small
The best model for data-intensive workloads.
Reason across millions of tokens in one prompt: entire repos, whole artifacts, and long-running agent state, with context to spare.
Learn MoreFrom the lab.
Rewriting the Scaling Laws of AI
Subquadratic is a frontier AI research company that believes the architecture layer is the highest-leverage surface in AI. While other major labs focus on incremental improvements to Transformer models, we're pushing foundational change at the architecture level to build models and products that scale efficiently.