affine

Affine is a decentralized model evaluation and coordination subnet designed to measure and reward meaningful improvements in AI model performance. Operating as Subnet 120 within the Bittensor ecosystem, Affine focuses on establishing shared evaluation standards that allow models to be compared under consistent and transparent conditions.

By providing a common reference point for performance, Subnet 120 helps reduce fragmentation across the network. Rather than relying on isolated benchmarks or subjective claims, Affine enables the ecosystem to identify which model changes deliver real, measurable gains, and to propagate those improvements more broadly within Bittensor.

Affine operates on a decentralized evaluation model that coordinates model testing across independent participants. As Subnet 120 within the Bittensor ecosystem, models are submitted to standardized evaluation environments where validators measure performance using agreed-upon tasks and metrics.

When a model demonstrates a clear and verifiable improvement over the current baseline, it is rewarded through Bittensor’s incentive mechanisms and may become a new reference for future evaluations. This process creates a continuous feedback loop where competition is based on demonstrable progress rather than incremental or cosmetic changes.