What is Space Bunny Alpha?
Space Bunny Alpha is a large language model on OpenRouter with a 1,000,000-token context window, native multimodal input, adjustable reasoning, and standout coding performance — now a named, paid model you build on. Here is what the Space Bunny model actually is, who is behind it, and how to use it in minutes.
The short version
Space Bunny Alpha (now the OpenRouter model minimax/minimax-m3) is an LLM that debuted on OpenRouter on 2026-09-23. It is a reasoning model: it produces hidden reasoning tokens before the visible answer, and you can dial the reasoning effort up or down. It accepts text, images, audio, and video as input and returns text. It has a million-token context window, is very fast, and is especially strong at coding. On OpenRouter it is now a paid, named model.
The stealth preview — and who made it
"Stealth" on OpenRouter means the real provider is hidden while a model is in preview. Labs use stealth launches to gather usage and feedback before attaching their brand. Space Bunny Alpha launched that way on 2026-09-23 with no named provider.
Community sleuths reverse-engineered the tokenizer and probed its behavior, and the leading identification was MiniMax M3. The stealth listing has since ended and the model is now served under MiniMax's named, paid id minimax/minimax-m3 on OpenRouter — matching that identification. This site is independent and is not the provider; we just make the model easy to try.
A 1,000,000-token context window
The headline capability is context. Space Bunny Alpha can hold up to 1,000,000 input tokens and emit up to 524,288 output tokens in a single call. That changes what you can do in one request:
- Load an entire repository and ask for a cross-file refactor or review.
- Feed a book-length document, contract, or spec and query it directly — no chunking or RAG plumbing.
- Keep long agent sessions coherent without aggressively summarizing history.
- Generate very large outputs — full files, long reports, big data transforms — in one shot.
Multimodal input, reasoning, and coding
Native multimodal
Send text, images, audio, and video in the same prompt. It reasons over all of them and returns text.
Adjustable reasoning
It thinks before it answers, producing hidden reasoning tokens. Turn effort up for hard problems or down for speed.
Strong at code
Fast inference plus a huge context make it a good fit for whole-repo coding agents and long refactors.
What does Space Bunny Alpha cost?
It graduated from the free stealth preview to the paid MiniMax M3 listing on OpenRouter ($0.30/1M input, $1.20/1M output). Our hosted endpoint is prepaid from $5 at $3 per 1M total tokens, so you only pay for what you run.
How to try the Space Bunny model
There are three easy ways to get your hands on it:
- Playground — open our playground, then grab a prepaid key to keep going.
- In your coding agent — wire it into a tool like OpenCode and let it work across your codebase.
- Hosted API — call our OpenAI-compatible endpoint at
https://spacebunnymodel.com/api/v1with a key likesb_live_.... Get an API key or read the usage guide.
curl https://spacebunnymodel.com/api/v1/chat/completions \
-H "Authorization: Bearer sb_live_..." \
-H "Content-Type: application/json" \
-d '{
"model": "space-bunny-alpha",
"messages": [{ "role": "user", "content": "Explain this repo." }]
}'Space Bunny Alpha specs
- Model id:
minimax/minimax-m3(OpenRouter);space-bunny-alphaon our endpoint - Availability: OpenRouter (now the paid minimax/minimax-m3 listing), since 2026-09-23
- Context window: 1,000,000 tokens
- Max output: 524,288 tokens
- Input modalities: text, image, audio, video
- Output: text
- Type: reasoning model, adjustable reasoning effort
- Strengths: coding, very fast inference
- Provider: anonymous (community guess: MiniMax M3, unconfirmed)
- Price: paid — $0.30/1M input, $1.20/1M output on OpenRouter