OpenCode launches Union Alpha model for free use on OpenRouter

Photo: Tima Miroshnichenko / Pexels

OpenCode launches Union Alpha model for free use on OpenRouter

The stealth coding model offers a 262,144-token context window, image support, and a zero-retention data policy during its free preview week.

A new AI coding model just appeared on OpenRouter with no price tag, no clear origin story, and capabilities that put it in the same conversation as frontier models. Union Alpha, a collaboration between OpenCode and OpenRouter, went live on September 16 as a “stealth model” designed for coding agents, complete with multimodal support and a context window large enough to swallow entire codebases.

The model is free to use during an initial preview period lasting roughly one week. After that, pricing details remain unclear. But for now, developers can experiment with what OpenCode describes as “frontier-level” general-purpose performance without spending a cent.

What Union Alpha actually does

The headline specs are genuinely impressive for a free offering. Union Alpha supports both text and image inputs, making it useful for tasks where developers need to feed screenshots, diagrams, or UI mockups alongside code. Its 262,144-token context window means it can process roughly 200,000 words of input at once, enough to analyze large repositories or lengthy documentation in a single pass.

Advertisement

The model is specifically optimized for agentic coding workflows, the kind of setup where an AI doesn’t just answer questions but actively writes, debugs, and iterates on code with minimal human intervention. OpenCode positions itself as a terminal-based AI coding agent, so Union Alpha slots in as the brain powering that workflow.

Perhaps the most notable policy detail: zero data retention. User prompts and model completions are not stored or used for training future models. OpenRouter’s pricing page lists the model at $0 with explicit zero-retention terms.

The stealth model playbook

Union Alpha isn’t the first model to appear on OpenRouter under mysterious circumstances. The platform has developed a pattern of hosting what it calls stealth models, AI systems released under the “Alpha” branding with minimal information about their actual developers.

OpenRouter functions as a routing layer, not a model developer. It connects users to various AI models through a unified API, but it doesn’t build the models itself. In the case of Union Alpha, the model is operated by an anonymous third-party provider. OpenRouter routes requests to it but makes no claims about who actually trained the thing.

This anonymity has fueled considerable speculation in developer communities. Some users have drawn connections between previous stealth model releases and Chinese AI labs, with names like Moonshot AI and Zhipu/GLM surfacing in discussions. None of these connections have been confirmed for Union Alpha, and the speculation remains exactly that.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
OpenCode launches Union Alpha model for free use on OpenRouter
OpenCode launches Union Alpha model for free use on OpenRouter

The stealth coding model offers a 262,144-token context window, image support, and a zero-retention data policy during its free preview week.

Photo: Tima Miroshnichenko / Pexels

A new AI coding model just appeared on OpenRouter with no price tag, no clear origin story, and capabilities that put it in the same conversation as frontier models. Union Alpha, a collaboration between OpenCode and OpenRouter, went live on September 16 as a “stealth model” designed for coding agents, complete with multimodal support and a context window large enough to swallow entire codebases.

The model is free to use during an initial preview period lasting roughly one week. After that, pricing details remain unclear. But for now, developers can experiment with what OpenCode describes as “frontier-level” general-purpose performance without spending a cent.

What Union Alpha actually does

The headline specs are genuinely impressive for a free offering. Union Alpha supports both text and image inputs, making it useful for tasks where developers need to feed screenshots, diagrams, or UI mockups alongside code. Its 262,144-token context window means it can process roughly 200,000 words of input at once, enough to analyze large repositories or lengthy documentation in a single pass.

Advertisement

The model is specifically optimized for agentic coding workflows, the kind of setup where an AI doesn’t just answer questions but actively writes, debugs, and iterates on code with minimal human intervention. OpenCode positions itself as a terminal-based AI coding agent, so Union Alpha slots in as the brain powering that workflow.

Perhaps the most notable policy detail: zero data retention. User prompts and model completions are not stored or used for training future models. OpenRouter’s pricing page lists the model at $0 with explicit zero-retention terms.

The stealth model playbook

Union Alpha isn’t the first model to appear on OpenRouter under mysterious circumstances. The platform has developed a pattern of hosting what it calls stealth models, AI systems released under the “Alpha” branding with minimal information about their actual developers.

OpenRouter functions as a routing layer, not a model developer. It connects users to various AI models through a unified API, but it doesn’t build the models itself. In the case of Union Alpha, the model is operated by an anonymous third-party provider. OpenRouter routes requests to it but makes no claims about who actually trained the thing.

This anonymity has fueled considerable speculation in developer communities. Some users have drawn connections between previous stealth model releases and Chinese AI labs, with names like Moonshot AI and Zhipu/GLM surfacing in discussions. None of these connections have been confirmed for Union Alpha, and the speculation remains exactly that.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.