Union Alpha revealed as Pareto 26.9, a multi-model system with no new training
A long post walks through the reveal: Pareto has several open and frontier models work the same task and check each other, calling a stronger model when needed; the team says there was no new training run, and the free preview ended after traffic hit billions of tokens per minute.
Union Alpha has been unmasked. The mystery model everyone was trying to identify is actually Pareto 26.9 from @TheUnbiasedCo. And the reveal is pretty interesting. Apparently, Pareto is a system that combines several existing AI models. Multiple open and frontier models work on the same task. Pareto checks their work and calls in a stronger model when needed. The team says there was no new training run. And they just needed to get real user feedback and test how it handled demand ahead of Pareto 26.10 next month. & holy shit, the demand was insane. According to the team, traffic hit billions of tokens per minute within hours. AWS even helped them triple their capacity overnight, and that still wasn’t enough. The free preview is now over, and Pareto 26.9 is available as a paid model. I find this pretty fascinating, especially considering how people were finding fingerprints pointing to GLM, Qwen, Llama and even Astra. Turns out, some of those fingerprints may have been coming from different familiar models working inside Pareto. Good tek.
This post quotes another post: x.com/unionalphaai/status/2100722366200557603