Models — UnboundModels

unbound/qwen3.8-27b

private beta · cold starts possible

Qwen3.8 27B Uncensored

A community uncensored build served in FP8 with a 131K context window: no provider-side moderation or hidden system prompt. The catalog links the published weights; API responses do not attest an exact checkpoint revision.

Weights: orcarouter/Qwen3.8-27B-Uncensored-FP8 ↗

Context131,072

Decode~92 tok/s

Input$0.35 / M

Repeated input$0.07 / M

Output$2.50 / M

Tool callsverified

Reasoningverified

Moderationnone

The repeated-input rate covers matching prefixes on the same account and model build within a rolling five minutes of a successful metered completion. Physical GPU cache hits are not required, even after restart; time to first token is not guaranteed.

Get an API key Read the quickstart →

unbound/frontier

in verification

Frontier

A larger uncensored build, still compact enough to serve per token. In the harness now. Same terms as everything else here: no provider-side moderation. Published weights and verified capabilities will appear here when it launches.

Specs and price land when it passes the harness suite.

The catalog grows only as fast as we can verify behavior. Want a specific model hosted? Tell us — requests shape what we test next.