Research

Publications & notes

Findings from Infinity Intelligence Labs, published as they land. The frontier model program has concluded; the research behind it has not.

2026-08-05 00:36

Introducing Xora 4 Flash: Local Intelligence in a Flash

Xora 4 Flash combines a Sparse MoE architecture with RWKV-inspired linear attention to deliver 256k-context reasoning, hosted entirely on-premise for absolute privacy and security.

Where the work continues

Labs is the research subdivision behind these publications, and the part of the division that carries on after the wind-down. Its current focus is efficiency, alternative architectures, context scaling and reasoning.

Infinity Intelligence Labs →