Crouton Alpha

OpenRouter added Crouton Alpha to its catalog on November 3 under the site's anonymous "Stealth" provider tag, describing it as a reasoning model built to handle the repeated sub-tasks AI agents run all day. It has a 1,048,576-token context window, supports up to 943,718 output tokens, and accepts text, files, images, audio, and video as input at zero price for both input and output. OpenRouter says the model is developed and operated by a third-party provider that has chosen to remain anonymous during the preview. That context size places Crouton Alpha in the same bracket as Google's million-token Gemini models, including Gemini 2.5 Pro, so a full codebase, legal filing, or hours of video transcript can be handled in one shot instead of chunked. The article notes OpenRouter has run this playbook since 2025, when Quasar Alpha and Optimus Alpha appeared free in April 2025 before being revealed as OpenAI GPT-4.1 tests.

Crouton Alpha appeared on OpenRouter on 3 November 2026 without a maker's name, and OpenRouter described it as a highly efficient reasoning model built to scale out the work agents repeat all day. The listing went live at 01:38 UTC, with the model ID stealth/crouton-alpha, a 1,048,576-token context window, and a maximum output of 943,718 tokens. Input covers text, image, file, audio, and video, while output is text only, with reasoning always on and adjustable from minimal to max, plus tool calling and structured outputs with tool choice limited to auto. Median speed is reported at about two seconds to first token and 72 tokens per second, pricing is $0 for input and output during the preview, and EU in-region routing is not available. There are no official benchmark scores, though one tester reported passing 140 of 157 tasks on his own set with a perfect coding score and weaker maths.

Crouton Alpha has been free to call since it appeared on 3 November 2026 at 01:38 UTC, which the analysis frames as a cash giveaway rather than the real story. Its provider field reads "Stealth," and the listing description indicates it is a preview of an unnamed model, keeping the anonymous-developer caveat intact. The economic framing compares call costs: a coding-agent task reading 200,000 tokens of repository and writing 20,000 tokens back costs four cents on GLM-5.3-Flash and twenty-three cents on Gemini 3.8 Flash. Filling the entire 1,048,576-token window with one prompt costs sixteen cents in input alone, or seventy-nine cents at Gemini's rate, illustrating the inference value of a free million-token window.

Crouton Alpha is a lightweight model built for scalable coding, agentic, and multimodal workflows, aimed at executing agentic sub-tasks, tool use, and document analysis in high-volume, repeated work. Developers can pair it with frontier reasoning models to run the bounded steps that convert plans and information into action. It was added to OpenRouter's catalog on November 3, 2026 under the anonymous Stealth provider tag. The listing shows a 1.05M context window with $0 per million input tokens and $0 per million output tokens during the preview period. OpenRouter states that Crouton Alpha is developed and operated by a third-party provider who has chosen to remain anonymous, and that OpenRouter only routes requests and is not the developer, owner, or provider. Prompts and completions may be retained by the provider but are not used for training, with other use governed by the Stealth Model Terms.