> ## Content Index
> Fetch the complete content index at: https://spyrigend.ghost.io/llms.txt
> Use this file to discover other available public pages before exploring further.

# Decentralized AI Gets a Boost from Multi-Subnet Collaboration on Quasar Models.
- URL: https://spyrigend.ghost.io/decentralized-ai-gets-a-boost-from-multi-subnet-collaboration-on-quasar-models/
- Published: 2026-06-29T21:25:33.000Z
- Updated: 2026-06-29T21:25:33.000Z
- Description: Bittensor Subnets 24, 56, and 3 are training long-context AI models together — a new cross-subnet collaboration push for decentralized AI infrastructure.
- Author: Quinn Hillerich
- Tags: All Articles, News, TAO

Quasar Models announced that its Subnet 56, which is part of the Gradients project, has joined Subnets 24 and 3 to train its long-context AI models on the Bittensor network. [Quasar Models](https://x.com/quasarmodels/status/2068741928041267288?ref=spyrigend.ghost.io) And so we have three subnets working together in a common training effort to push decentralized AI forward.

The partnership positions [Gradients](https://gradients.io/?ref=spyrigend.ghost.io) to play a major role in post-training and reinforcement learning. Such contributions are expected to transform base models into systems that are capable of chatting well, writing code, reasoning about complex problems, and performing better overall.

# Background on the major projects

[Quasar](https://subnetalpha.ai/subnet/quasar/?ref=spyrigend.ghost.io) is Bittensor Subnet 24 and is targeting long context foundation models. It trains and evaluates using a decentralized network of miners, which cuts costs relative to centralized approaches and transcends the constraints on how much context current AI models can handle. 

Gradients is Bittensor Subnet 56 — a competitive AutoML platform. Users upload datasets, select models, and miners compete to produce the highest performing versions via techniques such as instruction tuning and preference optimization. 

Subnet 3 is a component of this concerted training effort in conjunction with Bittensor’s larger ecosystem of decentralized compute and model development.

# Implications for decentralized AI

The partnership demonstrates increased cooperation between Bittensor subnets on common AI infrastructure. The groups hope to pool resources in training, post-training and specialized optimization to form an AI ecosystem that is free from centralization. 

This [development](https://github.com/SILX-LABS/QUASAR-SUBNET/?ref=spyrigend.ghost.io) is a good sign for further democratization and incentivization of model training through blockchain. The teams will be providing more updates as the project progresses.