--- format: "aidr-story-markdown/v1" id: "1599765a76b52d4e6a7ff0245950d521f01c12bda8818f161f40cdbf16a8bf36" canonical_url: "https://aidr.today/1599765a?lang=en" title: "Ox Alpha Frontier Model Debuts With 1M Token Window In Rare OpenRouter Stealth Launch" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-08-21T18:51:23.000Z" category: "Models" topics: ["llm","agent","open-source","coding","rag"] source_urls: ["https://x.com/alexatallah/status/2090828403486208361","https://x.com/opencode/status/2090758645499728234","https://x.com/cline/status/2090854216399220985","https://x.com/AskVenice/status/2090847078633156947","https://x.com/AndrewCurran_/status/2090795049080815978","https://x.com/Teknium/status/2090674052058984513","https://x.com/theo/status/2090649698705269157","https://x.com/teortaxesTex/status/2090652791018590552"] summary: "The newly released Ox Alpha AI supports multimodal text, image, and video inputs to handle sustained agentic work and production coding. It features a 1 million token context window designed for efficiency in real-world software deployment. OpenRouter distributed the model to partners including Venice, OpenCode Go, and Cline, which have opened free access to developers to gather testing feedback. No laboratory has claimed the identity of the stealth model, though the coding tool Cline reported early benchmarks showing marginal gains over Fable and GPT. OpenCode Go is providing near unlimited usage for 6 days, citing capacity for 100 trillion tokens per day to support the experimental release." --- # Ox Alpha Frontier Model Debuts With 1M Token Window In Rare OpenRouter Stealth Launch > [Open the canonical story]() **Published:** 2026-08-21T18:51:23.000Z **Category:** Models **Topics:** llm, agent, open\-source, coding, rag ## Summary The newly released Ox Alpha AI supports multimodal text, image, and video inputs to handle sustained agentic work and production coding\. It features a 1 million token context window designed for efficiency in real\-world software deployment\. OpenRouter distributed the model to partners including Venice, OpenCode Go, and Cline, which have opened free access to developers to gather testing feedback\. No laboratory has claimed the identity of the stealth model, though the coding tool Cline reported early benchmarks showing marginal gains over Fable and GPT\. OpenCode Go is providing near unlimited usage for 6 days, citing capacity for 100 trillion tokens per day to support the experimental release\. ## Sources - [Story source]() - [Supporting source]() - [Supporting source]() - [Supporting source]() - [Supporting source]() - [Story source]() - [Supporting source]() - [Supporting source]()