OpenRouter lists Ox Alpha as an anonymous stealth model and states that it is developed and operated by a third party that has chosen to remain anonymous. OpenRouter is only the routing layer. The page positions it as a reasoning model for coding, sustained agentic work, and production workloads, including workflows that use visual context.
Context window: 1,048,576 tokens (about 1M).
Maximum output: 131,072 tokens.
Positioning: coding, sustained agentic work, and production workloads.
Input modalities: the page mentions text and visual context; text, image, and video input are stated explicitly in OpenRouter’s official X post, covered in paired item 02.
Availability: free during the preview; the developer identity remains anonymous.
The page verifies that the mystery model exists and that it is positioned around long context, reasoning, agents, and visual context. It does not name GLM, Z.ai, or GLM-5.3-Flash. The editorially safe sequence is to present Ox Alpha’s verifiable specs first, then describe GLM-5.3-Flash as a community fingerprint attribution rather than an official OpenRouter or Z.ai name.
The page provides no standardized public benchmark table and does not disclose a formal model ID, parameter count, or weights. Google snippets repeat community claims such as “around 80% on DeepSWE,” but those numbers are not reproducible evidence from this page and should not be presented as benchmark facts.
GLM-5.3-Flash
