LLooMA v1.0 is a network-native LLM that's not defined by weights and parameters, but by the orchestration of behaviors across a distributed network.

LLooMA operates within the decentralized compute network PeerLLM. It takes a user prompt, analyzes it, decomposes it into tasks, and dynamically constructs a Directed Acyclic Graph (DAG) in real time. Each node in this graph maps to a physical host on the network, selected based on the models they have available.

LLooMA manages the full lifecycle of the response:
0/ Task distribution across hosts
1/ Response aggregation
2/ Anonymization and validation
3/ Formatting and streaming

The result is a single, unified response that is generated collaboratively across the network and delivered seamlessly to the user.

Explore the full architecture in the in-depth blog post:
https://blog.peerllm.com/peerllm/architecture/llooma/2026/04/11/llooma-in-depth.html

We Are .NET

LLooMA (LUMA) is a network-native behavior model built on the Pure LLM decentralized compute network. Unlike traditional LLMs, it acts as an orchestrator: it interprets user prompts, extracts intent, breaks requests into subtasks, generates a directed acyclic graph (DAG) of those tasks, routes them in parallel to multiple LLM-hosting nodes on the network, and aggregates the results back to the user in real time. A live demo shows a multi-part Japan travel planning prompt being distributed across six network hosts simultaneously, with each host handling a different subtask and streaming results back concurrently.

Introduction to Network-Native LLM (LLooMA)