Nuro AI Labs is a London-registered AI research lab founded in November 2024. It works across language models, architectures, reasoning, AI agents and supporting systems, with personal intelligence—persistent memory and per-agent cognition—as a unifying research focus. The lab publishes open research and releases open-weight models, while also developing commercial infrastructure and applied products.

Its research includes AVALON-2B, a sub-3-billion-parameter language model designed for self-reflective retrieval, available under the Apache 2.0 license; PLMR, a pre-tokenizer memory-routing approach for byte-level models; and Hydra, a next-generation architecture in development. Research materials, model weights and tools are distributed through platforms including Hugging Face, Ollama and GitHub.

Hypersave is the lab’s cognitive memory layer for AI agents, offering memory, retrieval and answer synthesis through Python and TypeScript SDKs. Applied products include Khyaa, personal intelligence for US financial advisors, and the Nuro stack, which includes Nuro Studio, Nuro Chat and Nuro One. Nuro AI Labs also develops solutions for memory-augmented agents, personal AI and sovereign AI environments.