Skip to content

Quasar Acquires Bittensor SN27 and Relaunches It as Orion Data Subnet

The SILX AI team says Orion will use Bittensor miners to generate and curate training data for Quasar, SN3 and a planned commercial data product.

Table of Contents

Quasar has acquired Bittensor Subnet 27 and introduced a new identity for the network: Orion.

Announced on July 12, Orion will focus on generating, discovering, and curating high-quality data for training the Quasar model series. The team also intends to make the subnet’s datasets available to other projects across Bittensor and, eventually, AI developers and companies outside the ecosystem.

The acquisition gives the SILX AI team two complementary pieces of decentralized AI infrastructure. Quasar’s existing SN24 coordinates the continued training and evaluation of its long-context foundation models, while Orion will work further down the stack to supply the data used during that training.

“Quasar builds the models. Orion provides the fuel,” the team said.

Why Quasar Is Building a Dedicated Data Subnet

Quasar said the idea for Orion emerged from a constraint the team encountered while training its models: access to enough high-quality data.

“Until now, we have trained using a combination of public datasets and a relatively small portion of high-quality synthetic data from Adaptive AI. However, to make Quasar truly SOTA, we need access to the best training data possible at a much larger scale.”

Producing that data through traditional methods can be expensive. According to Quasar, generating a dataset containing hundreds of billions of high-quality tokens can cost more than $100,000, depending on the models, filtering systems, and verification methods involved. So, instead of continuing to have data availability as an external limitation, Quasar decided to build a subnet around solving the problem.

The team believes Orion can reduce that cost to roughly $10,000 by distributing data discovery and generation across Bittensor miners. Each miner would contribute to finding or producing billions of tokens of potential training data, while Orion’s incentive and validation systems would be responsible for identifying the submissions most useful for model development.

Quasar said the idea for Orion developed through its collaboration with Bittensor SN3 and discussions with Bittensor co-founder Const.

The team also credited Mark Creaser and Siam Kidd of DSV Fund as investors and supporters who helped it secure the SN27 slot. Financial terms of the acquisition were not disclosed.

Bittensor Hedge Fund DSV Reportedly Raising $20 Million for TAO-Native Strategy
The Bittensor-focused fund is preserving flexibility while investing across the network’s emerging subnet economy.

Orion Moving Forward

Quasar does not intend for Orion to operate solely as an internal data pipeline.

The team is also developing a product built around Orion’s incentive mechanism that would supply AI training data to developers and companies outside Bittensor. The commercial strategy could create a feedback loop between the product and the subnet, where external demand would fund more data production, increased subnet activity would expand the available datasets, and better data could then support stronger Quasar models.

The announcement concluded with the team sharing they'd be introducing Orion-dedicated team members, while the core SILX AI team continues building Quasar and Orion in tandem.

To follow along with the journey from here, follow Orion on X - https://x.com/OrionSILX


Disclaimer: This article is for informational purposes only and does not constitute financial, investment, or trading advice. The information provided should not be interpreted as an endorsement of any digital asset, security, or investment strategy. Readers should conduct their own research and consult with a licensed financial professional before making any investment decisions. The publisher and its contributors are not responsible for any losses that may arise from reliance on the information presented.

Comments

Latest