Space Bunny is an anonymous reasoning model that emphasizes large context and adjustable reasoning capabilities. According to official materials, its core features are a 1 million token context window, support for text, image, and video inputs, and adjustable reasoning effort.
In terms of use, Space Bunny is aimed at practical work requiring long-context reasoning, suitable for code, research, visual understanding, and agent workflows, completed through a unified API. Its official description of uses includes reviewing code, reasoning across long documents, and asking questions about screenshots, charts, or videos and receiving text replies.
In terms of main capabilities: the context window is 1,000,000 tokens; it supports text, image, and video inputs; it provides five reasoning levels (low, medium, high, xhigh, max) to match the speed and depth of a task; it supports tool calling, where the model can request functions and return structured parameters; it provides a developer-friendly API using the familiar chat completions format, optional JSON responses, and standard usage metadata; it also provides an online playground for trial use.
In terms of usage, the official process is as follows: keep model calls on your own server, choose an appropriate reasoning effort, and send text or multimodal context in the standard chat format. The first step is to send a prompt, long document, source code, image URL, or compatible video URL, for example messages = [{ role: 'user', content }]; the second step is to set the reasoning effort; for everyday tasks you can start with low, and increase it when deeper reasoning is needed, for example reasoning = { effort: 'low' }; the third step is to use the response, where you can render the answer, request JSON, or connect tool calls, while the application retains control over side effects, for example answer = response.choices[0].message. In terms of the API, it is an OpenAI-compatible REST interface, supporting streaming output, JSON, and tool calling; the API key should be kept on the server side and should not be exposed in browser code or public prompts. When integrating with agents, use the Space Bunny endpoint and follow the chat request format already understood by existing clients, keep the API key in the server environment, expose only approved tools, and validate parameters and permissions before execution.
In terms of applicable scenarios, the scenarios officially listed include: in longwork in documents, large code repositories, and long sessions without having to chop tasks into pieces; ask questions about source code, screenshots, diagrams, or videos; give agents more room to think and choose reasoning intensity by task; connect products through the API to handle text, multimodal context, structured output, and tool calls.
Regarding limitations and boundaries, the official statement explicitly says that the model is anonymous and in preview, and its architecture, parameter count, training details, and knowledge cutoff are all undisclosed. Usage adopts a usage-based credit system, credits never expire, it is a one-time purchase, with no subscription and no automatic renewal.
Regarding pricing and plans, the official listing is: you can first try it for free in the playground, then choose a permanent credit pack; 100,000 permanent credits are suitable for experimentation, prototyping, and first integrations, including 100,000 credits, no expiration, 1 concurrent request, text/image/video prompts, text and JSON output, adjustable reasoning intensity, an online playground, and API key management; 1 million permanent credits are suitable for regular long-context, multimodal, and tool workflows, including 1 million credits, no expiration, unlimited workspaces, 3 concurrent requests, long-context conversations, usage and request history, and priority support; 11 million permanent credits (including a 10% bonus) are suitable for teams and high-load work, including 11 million credits, no expiration (including a 10% bonus), all Pro features, 10 concurrent requests, a dedicated fast lane, team workspaces and collaboration, custom integration support, security and permissions guidance, dedicated support, priority production processing, and product roadmap feedback.
Regarding other information, the official blog mentions articles on Space Bunny LLM's features, API, and evaluation guide, a practical evaluation guide for the Space Bunny Alpha model, a guide to access, limits, and setup for the Space Bunny free model, and the technical specifications, API, benchmarks, and real-world use cases of the Space Bunny Alpha AI model. The official FAQ entries cover what Space Bunny is, whether you can try it first, how much context is supported, what reasoning levels there are, what inputs and outputs are supported, who built the model, and how to contact the team, but the materials do not provide the specific answers to these questions.
