Open models.Unfiltered.
Uncensored open weights, served as published.
No added moderation. Pay per token.
Qwen3.8 27B Uncensored
131K context Tools & reasoning
- Input
- $0.35
- Repeated input
- $0.07
- Output
- $2.50
https://api.unboundmodels.com/v1Security research
Dissect exploit write-ups, malware reports, and attack chains.
OSINT
Cross-reference public reports, names, and timelines.
Red-teaming & evals
Stress-test prompts. Compare refusals and failure modes.
Fiction & long-form
Dark fiction. Complex characters. Room for the whole story.
Agent workflows
Plan multi-step tasks, call tools, and return structured results.
One model today. Deliberately.
We add a route only after it passes verification, so the catalog grows slowly and nothing in it is a surprise.
Qwen3.8 27B Uncensored
A community uncensored build served with a 131K context window. The published weights are linked below; API responses do not attest an exact checkpoint revision.
Your code already works.
Any OpenAI SDK, any framework, any agent harness — the entire migration is one line.
- 1Sign in with an email code and create an API key.
- 2Point
base_urlat our endpoint. - 3Top up a prepaid balance and pay per token. That’s all.
client = OpenAI( - base_url="https://api.openai.com/v1", + base_url="https://api.unboundmodels.com/v1", api_key=os.environ["UNBOUND_API_KEY"], )
Every promise here has something behind it.
Verified before it ships
A model reaches the catalog only after it runs the full harness suite. If something fails, we don’t advertise it.
- tool calling · multi-turnpass
- reasoning parserpass
- structured output · JSON schemapass
- 131K retrievalpass
A published model route
Use this model ID in your request. The catalog links the published weights; the API does not return checkpoint or revision metadata.
Request model ID { "model": "unbound/qwen3.8-27b" }
No prompt or completion storage
Message content is not written to disk. Billing uses usage metadata and short-lived keyed prefix fingerprints, not stored prompts or token IDs.
We store
- token counts
- timestamps
- model and key id
- short-lived keyed prefix fingerprints
We never store
- your prompts
- completions
- tool arguments
The model sleeps. You wake it when you need it.
Dedicated GPUs are expensive to keep idle, so we don’t — and we don’t hide it either. You start the model with one button in the console, or let an API key do it. While it starts, the stream stays open and tells you what is happening, so your client stays connected instead of timing out.
A start commits a $0.50 start credit that your tokens are paid from first. The GPU sleeps again after 5 minutes without requests. Physical prefix caching can speed up repeated requests, but cache residency and first-token latency are not guaranteed.
example request · streaming curl -N https://api.unboundmodels.com/v1/chat/completions \ -H "Authorization: Bearer $UNBOUND_API_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"unbound/qwen3.8-27b","messages":[{"role":"user","content":"Hello"}],"stream":true}' # While starting: SSE comments, ignored by SDKs : unbound-status allocating # Once ready: normal OpenAI chunks (choices excerpt) "choices":[{"index":0,"delta":{"content":"Hello"},"finish_reason":null}] # End of a successful stream data: [DONE]
Where else can you run this model?
We went looking. Every option we found, ours included.
| Option | Has an uncensored build | Context | Works with agents | Price |
|---|---|---|---|---|
| OpenRouter, Together,DeepInfra, Fireworks, Novita | no — official checkpoints only | — | — | — |
| Wiro AIhosts one | yes | 262K | own run API | not published |
| Featherlesscommunity re-uploads | ten of them | 32K | yes | $25–50 / mo |
| Self-hostingweights from Hugging Face | yes | 262K | you build it | a GPU |
| UnboundModelsone verified checkpoint | yes | 131K | yes — tools verified | per token · private beta |
Comparison snapshot: August 2026. Our model sleeps between sessions; you start it when you need it, with a $0.50 start credit that is spent on your tokens first.
Pay per token. Start the GPU when you need it.
No subscription and no idle fees. The model sleeps between sessions. You start it with one button, and the $0.50 you commit at the start is yours to spend on tokens.
- Input
- $0.35 / M
- Repeated input
- $0.07 / M
- Output
- $2.50 / M
Start credit: $0.50 per GPU start you ask for. Your tokens in that session are paid from it first; the part you have not used when the GPU goes back to sleep is not returned. No start credit when the model is already awake, and nothing is charged if a start fails.
Owner-approved private beta; public checkout is not open. Matching input prefixes on the same account and model build receive the repeated-input rate within a rolling five minutes of a successful metered completion, even after GPU cache eviction or restart. This is a billing discount, not a faster-first-token guarantee.
The example is a worked calculation at the rates above, not a measured session. The 12.6 M repeated tokens are eligible prefixes reused within five minutes; the 1.4 M regular input tokens are separate. Your totals depend on reuse and timing.
The questions you were going to ask anyway.
Is this legal?
Yes. These are published open-weight models, served under their own licenses. We are a hosting provider with published terms, operating from the EU.
So there are no rules at all?
There is no moderation layer between you and the model. There is still an acceptable-use policy: illegal use gets an account closed. Removing a filter does not remove the law.
Do you read or keep my prompts?
We do not store prompt or completion content or use it for training. Short-lived keyed prefix fingerprints support the five-minute repeated-input discount; see our privacy policy.
What happens when my balance runs out?
Requests stop until you top up. No overdraft, no invoice, no auto-charge to remember to cancel.
Why is there a start credit?
Waking a GPU costs us money before your first token. Starting it commits $0.50 of your balance, and your tokens in that session are paid from it first, so a real session pays nothing extra. If the start fails, nothing is charged.
What if I send a request while the model is asleep?
The API answers 409 model_asleep and charges nothing. Start the GPU in the console, or turn on auto-start for an API key so your agent can do it. A start takes a few minutes; the GPU sleeps again after 5 minutes without requests.
When do more models arrive?
As fast as verification allows. Tell us what you want hosted — requests shape the order we test in.
Full terms of use and the acceptable-use policy are short enough to read.