About SubQ

SubQ is developed by Subquadratic, a Miami-based AI infrastructure company founded by Justin Dangel and Alexander Whedon. SubQ is the first model built on a fully sub-quadratic sparse attention (SSA) architecture, meaning its compute cost grows linearly rather than quadratically as context length expands. This allows it to process up to 12 million tokens in a single pass — twelve times the context window of most top models. SubQ claims 56x faster prefill speed than dense attention models at 1M-token context, and achieves frontier-level coding performance at roughly 1/20th the cost of comparable models. The model is built for long-context tasks like reasoning across entire codebases, analyzing massive document sets, and maintaining persistent agent state without quality loss. SubQ has raised $29M in seed funding and offers API access, SubQ Code for developers, and SubQ Search for enterprise teams.

Key Features

Use Cases