Space Bunny Alpha › Free Tier & Limits
Is Space Bunny Alpha Free? Limits and Availability
Yes, Space Bunny Alpha is free right now — but it is not free in the way most guides describe, and that distinction is the whole point of this page. It is not a :free model variant: the listing itself is priced at zero. That changes which limits apply to you, and it means the "free tier" articles you have read about TokenRA probably describe a different mechanism. What follows is what is actually enforced, what is merely reported, and what disappears the day a vendor claims the model.
Last updated 25 September 2026 · Figures snapshotted 24 September 2026 · Model ID space-bunny-alpha
How we label sources on this page
Anonymous models attract a lot of confident numbers. We separate what TokenRA states from what a third party has measured from what is only a hypothesis.
Is Space Bunny Alpha Free? The Short Answer
| Question | Answer | Source |
|---|---|---|
| Is it free? | Yes — no charge for prompt or completion tokens | Provider-listed |
Is it a :free variant? | No — the ID does not end in :free | Provider-listed |
| Do free-variant request caps apply? | A separate mechanism — check your own key's counter | Provider-listed |
| Is a credit card needed? | No, while the preview is free | Provider-listed |
| Can a negative balance block it? | Yes — free models included | Provider-listed |
| How long will it stay free? | Unknown — historically until the model is claimed (about a week) | Independent |
"Free" Space Bunny Alpha vs a :free Variant
TokenRA has two distinct things that both look free from the outside, and the documentation draws a hard line between them.
A free variant is a model whose ID ends in :free — a free edition of a model that also exists in a paid form. TokenRA applies published free-usage request caps to these: a per-minute ceiling and a per-day ceiling, with the daily ceiling set by how many credits the account has purchased over its lifetime. Accounts with fewer credits get the lower daily figure; the higher ceiling is granted from one credit below the stated threshold to absorb rounding and top-up fees, which is why some accounts report is_free_tier: false together with the lower ceiling.
Space Bunny Alpha is not that. Its model ID is space-bunny-alpha, and it does not end in :free because there is no paid variant of it — the listing is simply priced at zero for the duration of the stealth preview. So the question "what is the daily request cap for this free model" is not answered by the free-variant table, and any guide that quotes you those numbers for this model is answering a different question.
The practical consequence: the free-variant caps are not your ceiling here, but they are also not your protection. What actually limits you is the upstream provider, and there is only one of those. See the availability section below — that is where requests actually fail.
How to Check Space Bunny Alpha Limits
Because tiers differ by account history, the authoritative answer for you is not a table — it is a call. GET https://tokenra.io/v1/key with your key returns the fields that matter:
| Field | What it tells you |
|---|---|
limit_remaining | Remaining credits on the key, or null if unlimited |
is_free_tier | Whether the account has ever purchased credits |
free_model_daily_requests.used / limit / remaining | The daily free-model counter and ceiling that apply to your account |
usage_daily / weekly / monthly | Credits consumed in the current UTC day, week (from Monday) and month |
Two caveats on reading that response. The free_model_daily_requests object reports the ceiling that gates your account where those limits apply — accounts and endpoints exempt from free-model limits, and BYOK requests, are not gated by it, so a non-zero remaining there reflects tier policy rather than an enforced ceiling. And the per-minute limit is not reported in that object at all; only the daily figure is. Monitor before requests start failing, not after.
The negative-balance trap
This one catches people precisely because the model is free, so the possibility does not occur to them. TokenRA's credit-limit documentation states that if an account has a negative credit balance you may see errors "including for free models," and that adding credits to bring the balance above zero restores access to them.
A zero-priced listing does not exempt an account from the balance check. If an expensive model on the same account has drawn the balance negative, a free model on that account can start failing with 402 — which reads like a bug until you know the rule. Both 402 and 429 are resolvable, but they need different responses: parse error.metadata.limit_source to tell budget exhaustion (tokenra_in_flight_budget, transient — honour Retry-After) from an exhausted key limit (tokenra_key_limit) from an insufficient balance (tokenra_credits, where retrying will not help).
Note also that the in-flight spending budget does not apply to requests to free models or to requests served entirely with your own provider keys. So on this model, budget holds are not the thing that will reject you — the upstream provider is.
Space Bunny Alpha Availability: The Number to Plan Around Listing-reported
The OpenRouter listing observed on 25 September 2026 reports uptime and availability as separate measurements, and conflating them is the most common misreading of these pages. OpenRouter defines uptime as the share of the past three days in which at least one provider responded. Availability is the share of the window in which inference was successfully served; OpenRouter says it uses the next-best provider when one returns an error. These definitions belong to the OpenRouter gateway snapshot and must not be used to infer the TokenRA route topology.
| Metric | Value (25 Sep 2026) | Reading |
|---|---|---|
| Uptime (3 days) | 99.76% | At least one provider was responding |
| Availability (3 days) | 98.24% | Inference was actually served |
| OpenRouter availability (24 hours) | 98.57% | Inference successfully served over the last 24 hours |
| Throughput | 80 tok/s (P50, best across providers) | Serving speed, not quality |
| Latency | 1.05 s (P50, best provider) | Round-trip time |
The snapshot shows a 1.52-percentage-point gap between uptime and three-day availability; the 98.24% availability figure means 1.76% of the measured window did not result in successfully served inference. OpenRouter explains that it continuously monitors providers and uses the next-best provider when one returns an error. These values therefore describe OpenRouter's gateway-level service metric, not an independent test of the anonymous upstream provider or TokenRA's routing.
What "free" costs you: retention and anonymity
The listing carries an explicit notice, and it should be read as a term of the deal rather than boilerplate: this stealth model is developed and operated by a third-party provider, and prompts and completions may be retained by that provider — while stating that they are not used for training, with all other use governed by the stealth model terms.
Combine that with the two facts that define a stealth listing — the operator is anonymous by design, and the arrangement is explicitly temporary — and the risk profile is not the same as calling a model whose owner you can name and whose data policy you can read. The honest framing: you are trading a small amount of data exposure and a real amount of availability for free inference at 1M context. That is often a good trade for evaluation and experimentation. It is a poor default for anything holding credentials, customer records, or personal data.
What Happens When Space Bunny Alpha Is Claimed
The free period has a known end mechanism. Stealth Alpha models exist to be revealed; the free window is the test period, and it closes when the vendor attaches its name to the result.
- Ox Alpha — free for roughly six days, then revealed by Z.ai as GLM-5.3-Flash.
- Union Alpha — revealed as Pareto by Unbiased; its free stealth period has now ended, and TokenRA's own listing directs users to the released model instead.
- Space Bunny Alpha — no claim as of 24 September 2026. All seven stealth Alpha models tracked since early 2026 have been claimed, most of them by Chinese labs, typically about a week after appearing.
Plan for that rather than being surprised by it. Two things are worth doing while it is free: pin the exact model slug in your config so the switch to a renamed model is a one-line change rather than a refactor, and record your own evaluations now, because the moment the vendor publishes it you will want your numbers beside theirs. Also worth noting: when a stealth model is claimed it may not become cheaper or more available — it becomes named, which is a different thing.
Space Bunny Alpha Free Limits FAQ
Is Space Bunny Alpha free?
Yes, as of 24 September 2026. The listing price is zero, so prompt and completion tokens are not charged during the preview.
Is it a :free model?
No. The model ID is space-bunny-alpha and does not end in :free, because it is a listing priced at zero rather than a free variant of a paid model. TokenRA's published free-variant request caps are a separate mechanism.
Are there daily request limits?
The published per-minute and per-day caps apply to :free variants. For this model, check GET /v1/key for the counter and ceiling that apply to your account, since tiers vary by credits purchased. The per-minute limit is not reported there.
Can a negative balance block a free model?
Yes. TokenRA documentation states a negative credit balance may cause errors including for free models, and that bringing the balance above zero restores access.
How reliable is it?
In the 25 September 2026 OpenRouter listing snapshot: 99.76% three-day uptime, 98.24% three-day availability and 98.57% OpenRouter availability over the last 24 hours. These are listing-reported gateway metrics, so use them as a snapshot rather than an independent provider benchmark.
What happens when it gets claimed?
Historically the free preview ends — Ox Alpha was free about six days before becoming GLM-5.3-Flash, and Union Alpha became Pareto with its free period now closed.