by Unbiased
Routing system from Unbiased that dispatches requests across several models with a checking harness; stealth name Union Alpha.
Updated EU status checked Last shown
Sources: creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Paro 26.9 is a model offering from Unbiased, previously tested anonymously as the stealth model "Union Alpha" on OpenRouter and OpenCode. According to a creator report (Matt Wolfe), it was later revealed to be a system that routes among several underlying models and uses a checking harness to verify outputs, rather than a single standalone model.
In one creator's own "Bucky bench", Paro 26.9 reportedly used under a third of the tokens of GPT6 Astra and finished faster with a similar result. This is a single informal benchmark and has not been independently verified here.
No official website, documentation, pricing, or licensing pages could be fetched. Details such as the underlying models, context window, supported modalities and exact release date are unconfirmed. Access has so far been reported via OpenRouter and OpenCode, which suggests API and coding-agent use.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Stealth model 'Union Alpha' on OpenRouter and OpenCode was revealed as Paro 26.9, a system routing among several models with a checking harness. In the creator's Bucky bench it used under a third of GPT6 Astra's tokens and finished faster with a similar result.
Closest neighbours by meaning, not just by category.
API · Not Diamond
Intelligent model router API that picks the model and reasoning effort per step for coding agents to cut inference cost.
Library · NVIDIA
NVIDIA open-source model router that selects a suitable LLM for each task to reduce cost.
Model
Open-weight non-autoregressive System 1 decision model returning calibrated probabilities in one forward pass across 100+ languages.
Model · OpenAI
OpenAI's GPT-6 Astra, a frontier language model used for reasoning, agentic work and weight interpretation research.
API · OpenAI
OpenAI high-speed inference tier that serves model output far faster than standard, reportedly around 14x.
Model · Microsoft
Microsoft's small decision-scoring model returning calibrated probabilities for routing, classification and rubric grading.
