Ox Alpha Launches on OpenRouter: 1M-Context AI Model Available in Free Preview
On August 20, 2026, an unannounced frontier AI model listed under the identifier stealth/ox-alpha appeared in the public model catalogs of OpenRouter and OpenCode.
The model is listed with a 1,048,576-token context window (~1.05M tokens), a maximum generation output limit of 131,072 tokens, and native multimodal support across text, image, and video inputs. During its initial rollout week, the endpoint is available across both platforms with zero per-token API fees.
No commercial AI laboratory or research organization has claimed ownership of the model, and its technical architecture and parameter scale remain officially undisclosed. OpenRouter lists the model under an anonymous provider classification, while OpenCode has integrated the endpoint into its Zen tier, promoting it for automated repository analysis and terminal coding tasks.
The release marks a growing industry pattern of deploying unbranded model checkpoints directly onto inference aggregator networks to evaluate real-world developer workloads before committing to formal launch announcements or commercial pricing schedules.
Developers can currently test the endpoint directly using standard OpenAI-compatible API configurations pointed at OpenRouter or through the OpenCode desktop interface.
