On September 16, Union Alpha, a stealth model, appeared on OpenRouter free of charge for both inputs and outputs. The model is pitched for research, coding, and AI agent tasks, but the developer and operator behind it have not yet disclosed their identity.
In terms of specifications, Union Alpha features a context window of approximately 262,000 tokens, supporting multimodal inputs and tool calling. According to OpenRouter, Union Alpha’s performance places it among the top tier of models, with scores close to several frontier models such as GPT 5.6 Sol, Fable 5, and Grok 4.6 on the DeepSWE benchmark, but at a vastly lower cost.
Nevertheless, this remains a general benchmark and is not enough to conclude that the model outperforms competitors in real-world workflows. For users, the more meaningful test remains its performance on familiar tasks and how much manual correction the output requires.