Space Bunny Free Guide: $0/$0 Anonymous Model on OpenRouter, 1M-Context Multimodal, Free Until 2026-10-05

Space Bunny (OpenRouter model ID stealth/space-bunny-alpha) is an anonymous large model that shipped to OpenRouter and OpenCode in September 2026. The hard numbers (verified against the OpenRouter API on 2026-10-03): $0 input / $0 output, a 1,000,000-token context window, 524,288 max output tokens, native text + image + video multimodal input, five adjustable reasoning tiers (low / medium / high / xhigh / max, default max, reasoning mandatory), plus tool-calling, Function Calling and JSON-Schema structured output. The model's expiration_date field is 2026-10-05, so this free window most likely closes on that date. The community rates it "roughly GLM 5.3 Flash but faster," and within 3 days of launch it hit #1 in daily calls on both OpenRouter and OpenCode. Bottom line up front: if you want a "1M-context + multimodal + agent-capable" model on a $0, no-credit-required plan, Space Bunny is the most concrete free offer currently on OpenRouter — but it is anonymous, unclaimed by any vendor, and the free window has a real expiry, so heavy use should happen before 10-05 or you'll be watching it flip to paid or delist.

Free Allowance & Model Specs at a Glance

Free Path: From Signup to a Working Call

1. Go to OpenRouter and sign up with email or Google;
2. Search stealth/space-bunny-alpha in the model list and confirm $0/$0 plus expiration_date: 2026-10-05 on the detail page;
3. Create an API Key on the Keys page for API use;
4. No code needed: just pick Space Bunny in OpenRouter's web Chat to test for free;
5. For API: set model to stealth/space-bunny-alpha, tune reasoning_effort (low for speed, max for accuracy), add tool calls, or feed image/video frames as multimodal input;
6. In OpenCode, select it as the backend model to run agentic workflows.

Key reminder: the free window ends 2026-10-05. After that, stealth/space-bunny-alpha most likely becomes pay-per-token or is delisted from OpenRouter — so if you want it free, finish your long-document, big-repo, and multimodal jobs before the deadline. Don't assume the $0 is permanent.

What You Can Run for Free (Capped by the Window + Limits)

• Blazing-fast inference: community-tested "roughly GLM 5.3 Flash but faster"; a full website in minutes;
• Coding / frontend: a portfolio site in minutes; a single-file 1,564-line GPU fluid-sim demo; completion rated "slightly better than GLM 5.3 Flash";
• 1M long context: 1,000,000 tokens swallows a whole book or large codebase without truncation;
• Multimodal input: text / image / video, understands screenshots, design mockups, and video frames;
• Agentic automation: Function Calling, Tool Choice, and JSON structured output let you wire it into multi-step tool chains;
• Reasoning trade-off: low tier for fast simple tasks, max tier for hard ones — dial speed vs. quality yourself.

None of the above costs extra (it is $0/$0 inside the window). What you actually burn is the free window + OpenRouter's free-model rate limits. Longer tasks, bigger context, and more multimodal input hit the limits faster; check the live remaining count via free_model_daily_requests.

Identity Guess: Who Is It Actually?

Space Bunny's parameter count, whether it is MoE, and its attention scheme are fully undisclosed. The most supported guess is MiniMax M3.1 (Flash), though MiniMax has not claimed it. The corroboration: MiniMax separately announced M3.1-Flash-Preview (MoE, inheriting the M3 base, ~428B total / ~23B active per token, 1M context, five reasoning tiers, ~280ms TTFT / 210ms at Low, ~126 tokens/s generation), which overlaps heavily with Space Bunny's config. In short, anonymous models are a common "run the numbers quietly before branding" play.

Competitor Free-Tier Comparison

(Competitor prices/specs from the OpenRouter API, checked 2026-10-03; treat live model pages as authoritative.)

Gotchas to Avoid

• The free window really does expire: expiration_date 2026-10-05; don't treat $0 as a standing allowance;
• Anonymous = no official backing: architecture and identity are unconfirmed; uptime (~99.9%) is OpenRouter's live figure, so don't run production-critical paths on it;
• Free-model rate limits: OpenRouter rate-limits free models by all-time credits; the daily remaining free count is in free_model_daily_requests — bursts will hit 429;
• Reasoning is mandatory: default is max; dial reasoning_effort down to low to save time instead of burning the window on max.

FAQ

Bottom Line

Space Bunny's free policy is refreshingly numeric: $0 input / $0 output + 1,000,000-token context + 524,288 max output + text/image/video multimodal + low~max reasoning tiers + tool-calling, free window to 2026-10-05. If you want a "1M-context + multimodal + agent-capable" model on a $0 plan, it is the most concrete free offer on OpenRouter right now; anonymous, unclaimed, and with a real expiry — so finish your jobs before 10-05 and see whether it turns paid or delists. Pricing, limits, and listing status are all in flux; verify on the live OpenRouter model page before you commit.