Quasar Revolutionizes AI Training with Decentralized Marketplace

Quasar Models has built a decentralized AI training marketplace on Bittensor, aiming to reduce the cost of training competitive AI models. The project, developed by SILX AI under the SILX Labs umbrella, allows independent miners to operate and refine AI models through a competitive evaluation system.

Miners who improve model performance get rewarded, while those who don't are outcompeted. Quasar is targeting long-context foundation models with context lengths of approximately 2 million tokens or more.

The team has deployed the Quasar-3B architecture, which launched in April 2026 and claims a 99.5% reduction in pre-training costs compared to traditional centralized methods.

Quasar plans to launch a 10-trillion-token decentralized training run, split into two phases of 5 trillion tokens each. This ambitious project aims to put Quasar in the conversation with the largest training efforts ever attempted.