Via bitget.com
The decentralized AI protocol now supports multi-label predictions, moving beyond single scalar values for network outputs.
Allora Network pushed its mainnet to version 0.17 on August 19, upgrading its decentralized AI prediction protocol with a meaningful shift in how the network returns results. Instead of spitting out a single number, the system can now return labeled outcomes, a change that opens the door to categorization-style predictions rather than just point estimates.
What actually changed #
The core technical shift in v0.17 is straightforward but significant. Previously, Allora’s network of machine-learning models would aggregate their inputs and produce a single scalar value as output. With labeled outcomes, predictions can now carry semantic meaning alongside numerical values. A prediction about weather isn’t just “0.73”; it can be “rain: 0.73, snow: 0.15, clear: 0.12.”
The upgrade doesn’t break anything that already exists. Topics currently running on single-output predictions maintain backward compatibility through the migration path. The release also bumps the direct emissions integrations to version 10, refining how the protocol handles token distribution mechanics tied to network participation.
How Allora’s AI layer works #
Allora operates as a Layer 1 blockchain built on the Cosmos SDK. The network functions as what the team calls a “self-improving intelligence layer,” coordinating independent machine-learning models that compete and collaborate to produce better predictions over time.
The mechanism relies on dynamic aggregation. Multiple independent models submit their predictions, and the protocol weighs and combines them based on historical accuracy. Models that perform well get more influence. Models that don’t get less. Anyone can contribute a model and earn rewards through the network’s native ALLO token, which launched alongside the mainnet in November 2025.
Earlier in August, the team had already shipped updates related to worker promotion through statistical tests, creating a more rigorous vetting process for which models earn the right to contribute predictions on active topics. The v0.17 release builds on that quality-control theme by expanding what those vetted models can actually express.
What this means for the ALLO token and network growth #
The update to direct emissions integrations in version 10 matters for tokenomics. How a protocol distributes rewards to participants shapes the economic incentives for model contributors, validators, and other network actors.
No major partnerships or token unlock events accompanied the v0.17 release. Allora has been shipping upgrades at a steady clip since its November 2025 mainnet launch, with each version building incrementally on the last.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our