Crouton Alpha Free

The Sumus tracker records Crouton Alpha as an anonymous stealth model listed on November 3, 2026 under the ID stealth/crouton-alpha with a free preview price of $0 per million input and output tokens. It confirms a 1,048,576-token context window, text/image/video/file/audio input, text output with tool choice and JSON-schema support, and roughly 150 tokens per second at 2.27s P50 latency. The page separates confirmed facts from rumor, noting that OpenRouter says a third-party provider runs the model and chose anonymity during the preview, while guesses such as Meta or Google are unconfirmed. It also flags that prompts may be kept by the provider and are not used for training, and that the model is unavailable through EU in-region routing.

i-Scoop's November 4, 2026 report describes Crouton Alpha as a free anonymous AI model that appeared on OpenRouter on November 3, 2026 without a maker's name attached. It reads text, images, audio, video and files, has a context window of just over one million tokens, and is pitched at the small repeated steps inside AI agents, with reasoning always on and adjustable from minimal to max. The report notes the listing specifies about two seconds to first token and 72 tokens per second median, no EU in-region routing, and no official benchmark scores yet, citing one tester's 140 of 157 result with a perfect coding score. Usage was climbing quickly: on November 4 the model ranked second in OpenRouter's new and trending models with 25.1 billion tokens processed.

Crouton Alpha first appeared on a model aggregator's listing on 3 November 2026 at 01:38 UTC, and the listing's provider field reads "Stealth" rather than naming a developer, indicating an unnamed upstream model exposed through an inference endpoint. The post frames the release as a preview of an unidentified model rather than a named product launch. The same write-up describes Crouton Alpha as free to call with a 1,048,576-token context window, and quantifies the cost of filling that window with input alone at about sixteen cents. It also frames the model's appeal as a low-cost option for large-context coding-agent workloads that read large repositories and write back sizable patches.

AI/ML API's hands-on report on the stealth model Crouton Alpha, released November 3, 2026 by an anonymous provider, records a 1,048,576-token context window, advertised maximum completion of 943,718 tokens including reasoning, text/image/video/file input with inconsistent audio acceptance across endpoints, and text-only output. Reasoning is always on with six effort levels from minimal to max, with max as the default. In tests run on November 4, 2026, the model passed all nine of the publisher's coding, tool and log tasks, had the lowest median completion time across six short coding tasks compared with GLM 5.3 Flash and MiniMax M3.1 Flash, and passed 140 of 157 tests in the SMF Works Cold Iron run, two fewer than Claude Fable 5.1. The report also notes tool_choice "auto" was the only working setting, with structured JSON output supported via response_format and JSON Schema.

OpenRouter's model page lists Crouton Alpha (model ID stealth/crouton-alpha) as a stealth model released November 3, 2026, described as a lightweight model for scalable coding, agentic and multimodal workflows, ideally paired with frontier reasoning models for bounded agent sub-tasks, tool use and document analysis. The listing states the model is developed and operated by an anonymous third-party provider, and that OpenRouter routes requests without being its developer, owner or provider. Specifications on that listing show a 1.0M context window, free input and output pricing during the preview, and provider-side retention of prompts and completions without training use, governed by Stealth Model Terms. Performance telemetry on the page shows roughly 3.64s latency, 35 tokens per second and 97.65% uptime from the single hosting provider, with all request volume routed directly to it.