Home / Guides / Space Bunny

Published September 26, 2026 · routes and prices checked September 26, 2026 · applies to pgup-ai/jbot-review-action v0

Space Bunny is an anonymous model with a 1M-token context window, listed at $0 on OpenCode Zen, Kilo, OpenRouter and Cline since 2026-09-23. OpenCode Go ($10 a month) and Command Code (from $1 a month) also include it at no per-token cost. It’s the free model we’d start J-Bot Review on today. At the default high reasoning effort it found 6.5 of 27 known issues on our test pull requests, with a median review time of 251 seconds.

OpenCode setup in three steps

1. Create an OpenCode API key. Save it as the repository secret OPENCODE_API_KEY (Settings → Secrets and variables → Actions).
2. Commit the workflow. Add the file below as .github/workflows/jbot-review.yml. The opencode/ prefix selects the provider, and J-Bot already runs Space Bunny at high reasoning effort on this route, so nothing else needs setting.
3. Open a pull request. J-Bot reads the base…head diff and posts diff-anchored findings with a verdict. A second session checks blocking findings before they post.

name: J-Bot Code Review
on:
  pull_request: { types: [opened, reopened, ready_for_review, synchronize] }

concurrency:
  group: jbot-review-${{ github.event.pull_request.number }}
  cancel-in-progress: true

permissions:
  contents: read
  pull-requests: write
  issues: write
  checks: read

jobs:
  review:
    if: github.event.pull_request.draft == false
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v7
        with: { fetch-depth: 0 }
      - uses: pgup-ai/jbot-review-action@v0
        with:
          model: opencode/space-bunny-free
          opencode-api-key: ${{ secrets.OPENCODE_API_KEY }}
          github-token: ${{ secrets.GITHUB_TOKEN }}

Six routes, six exact model ids

Pick the route where you already manage credentials. The first segment of each id selects the provider. Prices are per million tokens from each live catalog on 2026-09-26, and quotas stay under each provider’s control.

J-Bot’s high-effort default for Space Bunny covers the two OpenCode routes and Command Code. On OpenRouter it runs at the global low default unless you set model-options. Kilo and Cline run through their own CLIs, which don’t take J-Bot’s effort option. Command Code charges no credits for Space Bunny on any plan, and its cheapest plan is $1 a month. J-Bot still checks your plan’s 5-hour, weekly and monthly limits before each run, and stops if no key has room.

Reasoning effort is the setting that matters

We ran Space Bunny on four pull requests from a production TypeScript monorepo, with 27 issues that developers had already accepted, two runs per setting:

• Low: 2.5 known issues, 68 s median review. Space Bunny reasoned for about a thousand tokens per session and posted about one finding.
• Medium: 3.0 known issues, 94 s.
• High: 6.5 known issues, 251 s.

High effort found more than twice as many issues for 3.7 times the wall time, so J-Bot made it the default on OpenCode and Command Code. Verification runs one tier lower, at medium. If you’d rather have a review in about a minute and accept fewer findings, set low effort yourself:

      - uses: pgup-ai/jbot-review-action@v0
        with:
          model: opencode/space-bunny-free
          model-options: '{"reasoningEffort":"low"}'
          opencode-api-key: ${{ secrets.OPENCODE_API_KEY }}
          github-token: ${{ secrets.GITHUB_TOKEN }}

Why effort helps, and what else we tested, is in How do you get better code reviews from free AI models?

What we know about Space Bunny

• Context. OpenCode lists a 1,048,576-token window with up to 524,288 tokens of input per request and 524,288 of output. OpenRouter and Kilo list 1,000,000.
• Tools and reasoning. Every catalog marks tool calling and adjustable reasoning. OpenRouter describes it as a model with fast inference, strong coding and native multimodal input.
• Speed. At low effort it was the fastest free model we’ve run in J-Bot. On the same test pull requests it finished in about 50 seconds where MiMo V2.6 Flash’s free route took about 24 minutes.
• Origin. Unknown. The catalogs call it a stealth or anonymous model, and none names a lab or a release plan.

Where the diff goes

• An unnamed lab serves the model. OpenRouter says the provider retains prompts and completions but doesn’t use them for training. The other listings state no policy, so treat what you send there as possibly kept and used.
• Cline’s free tier adds its own terms. Cline’s documentation says free-model usage may be used to improve model performance and quality.
• Your runner stays in control. On the OpenCode route the review session runs read-only on your GitHub Actions runner, and J-Bot sends the diff only to the provider you configure.
• On Cline, every tool is approved. Cline works in the checkout in plan mode with its shell and write tools auto-approved, and a checkout’s .cline/hooks and .clinerules load. Its tools include web search, so the model can send search queries of its own. J-Bot documents this as an accepted risk on CI runners, so use the Cline route on repositories where you trust the people opening pull requests.
• Fork pull requests can’t read the key. GitHub doesn’t pass repository secrets to pull_request workflows from forks.

FAQ

Related