SubQ AI

SubQ is an AI model platform focused on long-context work. It is built on a fully sub-quadratic architecture and is designed to handle very large inputs, including up to 12 million tokens, while keeping speed and cost efficient. The site says it is made for tasks like full codebase analysis, long-running agent state, and large-scale retrieval without quality loss. SubQ also offers early access through an API and a coding agent called SubQ Code.

Key Features:

• 12M-token reasoning for very large context tasks.
• Fully sub-quadratic sparse-attention architecture for lower compute use.
• OpenAI-compatible API endpoints with streaming and tool use.
• SubQ Code for working across entire repositories in one context window.

Use Cases:

• Analyzing large codebases from start to finish in one pass.
• Handling long research, document, or conversation histories.
• Powering coding agents that need broad project context and persistent state.