An anonymous model known as Ox Alpha, dubbed "Niu Lai" within Chinese tech circles, has seen its usage metrics skyrocket shortly after debuting on the OpenRouter platform, climbing rapidly to the top spot and setting a fresh record for single-day consumption. According to the latest data from OpenRouter, cumulative calls from just the top five applications, including Hermes Agent and Claude Code, have already surpassed 4 trillion tokens, with the buzz spreading swiftly from early adopters to real-world workflows used by developers and agents worldwide.

The model reportedly supports a context window of around 1 million tokens and handles text, image, and video inputs, earning strong praise from international users. Stripe CEO Patrick Collison took to social media to call it "quite impressive." In the latest full benchmark run, Ox Alpha scored 63 points on the 113-task DeepSWE test, placing its overall performance near that of Claude Opus 4.8.

Despite its growing fame, the maker of Ox Alpha remains unidentified. Technical fingerprints point more toward Zhipu's GLM series, yet several researchers from Google DeepMind have posted hints suggesting it could be a new iteration of Gemini. As of now, no company has formally laid claim to the model. Notably, OpenCode has announced that it has prepared 100 trillion tokens of daily service capacity for Ox Alpha. If the model is eventually confirmed to come from a Chinese developer, it would validate the ability of domestic large models to directly enter the global developer ecosystem, manage worldwide distribution, and handle inference supply, potentially boosting market interest in local models, AI inference infrastructure, and the broader agent application supply chain. The free access window for Ox Alpha is expected to close around August 26.