Space Bunny Alpha in opencode

opencode is the open-source coding agent everyone's pairing with Space Bunny Alpha right now — a fast, 1M-context model you reach through our hosted API or OpenRouter. Here's the exact config for both routes.

opencode lets you plug in any OpenAI-compatible model, and Space Bunny Alpha is one of the most capable options available right now: a 1,000,000-token context window, strong coding, and blazing-fast inference. That combination — huge context, fast, good at code — is exactly why it's trending in the opencode community. Below are two setups: the quickest (OpenRouter) and the most convenient (our hosted endpoint, no OpenRouter account and one less rate limit to worry about).

Option A — via OpenRouter (quickest)

opencode has built-in OpenRouter support. Log in with an OpenRouter key, then select the Space Bunny model (now the paid MiniMax M3 listing).

# 1. authenticate opencode with OpenRouter
opencode auth login
#   → choose "OpenRouter", paste your OpenRouter API key

# 2. run opencode and pick the model
opencode
/models
#   → search: minimax-m3  → select openrouter/minimax/minimax-m3

Or pin it in your opencode config so every session starts on Space Bunny:

{
  "$schema": "https://opencode.ai/config.json",
  "model": "openrouter/minimax/minimax-m3"
}

That's it — you're coding on a 1M-context model. Note that OpenRouter bills MiniMax M3 per token ($0.30/1M input, $1.20/1M output) and sets its own rate limits; Option B (our hosted endpoint) gives you one stable key and model name instead.

Option B — via our hosted endpoint (no OpenRouter account)

Prefer one key that just works — no OpenRouter signup, and a stable model name we keep pointed at the live Space Bunny even if the upstream id changes? Add our OpenAI-compatible endpoint as a custom provider. Grab an sb_live_ key on the API page first.

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "spacebunny": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "Space Bunny",
      "options": {
        "baseURL": "https://spacebunnymodel.com/api/v1",
        "apiKey": "{env:SPACE_BUNNY_API_KEY}"
      },
      "models": {
        "space-bunny-alpha": { "name": "Space Bunny Alpha" }
      }
    }
  },
  "model": "spacebunny/space-bunny-alpha"
}

# put your key in the environment (or opencode's auth store)
export SPACE_BUNNY_API_KEY=sb_live_...

Now opencode routes every request to spacebunnymodel.com/api/v1/chat/completions — the same OpenAI-compatible surface as OpenRouter, but on one key with a stable model name. Restart opencode and Space Bunny is in your model list.

Getting the best out of it in opencode

• Lean on the 1M context: point opencode at the whole repo instead of a handful of files — Space Bunny can hold it.
• It's a reasoning model: it spends output tokens thinking before answering, so give it room (a higher max output) on hard tasks.
• Billed per token on either route — predictable prepaid credits from $5 keep high-volume agent loops in check.
• Multimodal input (text, image, audio, video) is supported upstream if your opencode workflow sends it.

Which route should you pick?

Use Option A if you already have an OpenRouter key and just want to try Space Bunny in five seconds. Use Option B if you want a signup-free key, a stable model name that won't break if the upstream id changes, and higher, more predictable limits. Both hit the same model — the only difference is who sits in front of it.

FAQ

See also: Get a free API key · OpenRouter setup · How to call the API