Bittensor Co-Founder Outlines Strategy for Inter-Subnet Payments and Launch of Afine 1 AI Model

Bittensor co-founder Jacob "Const" Steeves outlined major protocol milestones and upcoming architectural updates during a keynote address at the network's Exploit conference. Highlighting the growth of the decentralized incentive computing platform—which now features 128 active subnets, 23 of which generate direct commercial revenue—Steeves unveiled "Aphine 1," a 36-billion-parameter artificial intelligence model trained entirely on Bittensor's decentralized infrastructure. Steeves stated that the model matches benchmark scores of established commercial systems, such as Anthropic's Claude 4.1 Opus, while reaching processing speeds of up to 500 tokens per second on H100 GPUs.

Alongside Afine 1, Steeves showcased ecosystem advancements including the 110-billion-parameter Tutonic 2 foundation model and the Albdo subnet's LLM-driven evaluation systems. To support these increasingly complex AI workflows, Bittensor is implementing protocol changes aimed at deepening subnet sustainability and interoperability. Key updates planned for 2026 and 2027 include canonical bridges designed to track external capital inflows, allowing subnets to earn higher network emissions by generating real-world revenue from outside buyers.

Introducing Gamma Tokens for Inter-Subnet Infrastructure

A major pillar of the network roadmap is the launch of "gamma tokens," a standardized format for usage credits across Bittensor infrastructure providers such as Liam, Targon, and Hypius. This system will enable subnets to allocate a portion of their block emissions toward purchasing compute, storage, or inference directly from peer subnets. Steeves emphasized that establishing these self-sustaining feedback loops within an open incentive mechanism ensures the network remains an autonomous, censorship-resistant computational layer capable of competing with centralized artificial intelligence laboratories.