Ox Alpha is a stealth reasoning model designed for coding, sustained agentic work, and production workloads. Developed by an anonymous third-party provider, it excels at long-horizon software engineering, complex reasoning, and multimodal workflows combining text with visual context. OpenRouter routes requests directly to the model via a unified, OpenAI-compatible API using the model ID `stealth/ox-alpha`.

The model offers a 1,048,576-token context window with up to 131,072 output tokens, supporting text, image, and video inputs. Performance metrics show 69 tps throughput (P50) and 1.00s latency, with 99.99% uptime over three days and 99.94% availability. Pricing is free for both prompt and completion tokens during this preview.

Ox Alpha supports tool calling, structured JSON outputs, and function calling via `tools` and `response_format` parameters. Released on August 20, 2026, it is currently served by a single provider, Stealth, with OpenRouter handling load balancing and error recovery. Top applications driving traffic include omp, Hermes Agent, and Claude Code, indicating strong adoption for agentic coding workflows.