Subquadratic 1.1 Small

#1 v Embeddings a vektorové DB

Subquadratic · v1.1 · small · 2× · naposledy 28. 9. 2026

100

Momentum

SubQ 1.1 Small is a Long-Context language model from Subquadratic (Miami) based on the proprietary Subquadratic Sparse Attention (SSA) – a content-aware sparse Attention that scales linearly rather than quadratically with context length. The model was primarily trained on 1M Token but generalizes to near-perfect Retrieval results up to 12M Token in the Needle-in-a-Haystack test, according to the Technical Report. It builds on an existing Open Source frontier model whose dense Attention has been replaced with SSA, and is currently available through limited access to selected design partners or via waitlist. According to third-party sources, no public pricing list exists yet.

Zdroje (2)