Lyber
Request access

Abliterated model router · Early access

Agentic freedom is human freedom.

AI should serve the user, not the companies.

Uncensored and abliterated Qwen, DeepSeek, GLM and Kimi models behind one OpenAI-compatible API.

Get startedRead the manifesto
  • OpenAI-compatible API
  • Five uncensored models
  • Prepaid, per token
  • lyber/qwen38-27b-abliteratedAbliterated
  • lyber/deepseek41-flash-abliteratedUncensored
  • lyber/glm53-full-fp8-abliteratedUncensored
  • lyber/glm53-flash-abliteratedUncensored
  • lyber/kimi-k3-abliteratedAbliterated

AI should serve you. Not the companies.

Right now, we are creating a level of intelligence that has never been seen before, and yet we are falling short at making it humanity-centric.

Don’t ask for permission.

The tools that expand what people can research, build and imagine should not wait on a policy team’s mood. Pick a model, send the request, read the answer.

Your intent is the spec.

An agent that serves you does what you asked, not a softened, rewritten version of it. You judge the answer, not a filter.

Your request

“Answer what I asked.”

model: lyber/auto

  • Qwen 3.8 27Blyber/qwen38-27b-abliterated
  • DeepSeek V4.1 Flashlyber/deepseek41-flash-abliterated
  • GLM 5.3 · Fulllyber/glm53-full-fp8-abliterated
  • GLM 5.3 · Flashlyber/glm53-flash-abliterated
  • Kimi K3lyber/kimi-k3-abliterated

The response names its model"model": "lyber/glm53-full-fp8-abliterated"

Illustration. lyber/auto routes each request to an eligible, qualified model, and the response names it. Inference opens after qualification.

One API. Your models. Your rules.

Get startedRead the manifesto

Hack the planet. Liberate the planet. Lyber.

The point

What we build

An obliterated router.

Diagram: your app sends an OpenAI-compatible request with one API key to lyber/auto, which routes it to one of five uncensored models (Qwen 3.8 27B, DeepSeek V4.1 Flash, GLM 5.3 Full, GLM 5.3 Flash, Kimi K3); the response names the model that served it.lyber/autoQWEN 3.8 27B1× A100 80GB · $0.30 / $1.50DEEPSEEK V4.1 FLASHmixed FP8/FP4 · $0.30 / $1.20GLM 5.3 · FULL8× H200 FP8 candidate · $1.50 / $4.00GLM 5.3 · FLASH4× H200 candidate · $0.14 / $0.45KIMI K38× B300 · $3.25 / $15.00YOUR APPOPENAI SDKCODEX CLICURLOPENAI-COMPATIBLE"model": "lyber/qwen38-27b-abliterated""model": "lyber/deepseek41-flash-abliterated""model": "lyber/glm53-full-fp8-abliterated""model": "lyber/glm53-flash-abliterated""model": "lyber/kimi-k3-abliterated"
Illustration. lyber/auto routes each request to an eligible, qualified model, and the response names it. Hardware is each model's launch plan; inference opens after qualification.

  • One API, powerful models.

    Censorship-free models behind one OpenAI-compatible endpoint. Point the SDK you already use at it and keep your code.

    Connect a client

  • Choose the model.

    Pick a named checkpoint or send to lyber/auto. Every answer tells you which model produced it.

    See the models

  • Your key is your capability.

    Requests run on one API key and prepaid dollar credit, topped up with a card or crypto. A website login never stands in for the key.

    Get API access

Monthly plans

Choose your plan.

Monthly credit. One balance for your browser and agents.

Browser · Codex · Hermes · API

Model ratesToken calculator →

Launch preview · not on sale

No charge to request access. Inference is closed.

Starter

$15/ month

$20 monthly usage credit

Benefits — Starter
  • $5 monthly bonus included
  • All available models & API
  • Supported browser & agent workflows
  • Credit expires at renewal
Learn more — Starter

Pro

Recommended

$45/ month

$60 monthly usage credit

Benefits — Pro
  • $15 monthly bonus included
  • 2-cycle banking · $120 cap
  • Team workspaces & usage history
  • Spending controls
Learn more — Pro

Max

Best value

$180/ month

$250 monthly usage credit

Benefits — Max
  • $70 monthly bonus included
  • 3-cycle banking · $750 cap
  • All available models & API
  • Team workspaces & spending controls
Learn more — Max

Enterprise

Custom

Usage and capacity by agreement

Benefits — Enterprise
  • Custom volume pricing
  • Agreed rate & concurrency limits
  • Hosting by agreement
  • Custom support terms
Learn more — Enterprise
Plan details

Every plan includes: Available models · Browser chat & API · Supported Codex & Hermes workflows · Key backups · Usage history · Team workspaces · Spending controls

Same model token rates on every plan. USD before tax. Client support depends on the qualified model and integration.

One-time credit from $10. Pay by card, with no subscription. Purchased credit stays separate from expiring plan credit.

How does credit work?

Starter: $15 for $20 credit. Pro: $45 for $60. Max: $180 for $250. Spending that included credit is not an additional charge; requests use the selected model’s token rates. Extra prepaid credit is purchased separately.

What happens when my monthly credit runs out?

Choose to add prepaid credit or wait for renewal. The plan design does not authorize automatic overage charges. Automatic top-ups would require your explicit opt-in and a spending cap.

Where can I use my credit?

Across your Lyber account: browser, API and supported Codex or Hermes workflows. Compatibility depends on the client and qualified model. A higher plan adds credit and banking, not model access, context or features.

Will unused monthly credit roll over?

In this proposal, unused Starter credit does not roll over. Pro credit stays usable through the following billing cycle, capped at $120; Max credit lasts through two following cycles, capped at $750. Oldest plan credit is spent first. Separately purchased prepaid credit does not become expiring plan credit. Renewal, cancellation, expiry and refunds must be implemented and final terms published before these benefits are sold.

Are there usage or capacity limits?

Credit buys metered tokens, not unlimited inference. Fresh input, recognized cache reads and output have separate rates; billable reasoning counts as output once. Plans do not reserve hardware or promise rate limits, queue priority or support response times. Final limits and terms will be published before subscriptions open.

Are subscriptions available now?

No. These are proposed prices and benefits. Requesting access creates no charge or subscription and does not enable inference. Final terms, including renewal, cancellation and refunds, will be published before sale.

Add credit from $10

Give $15. Get $15.

$5 signup credit now · paid bonuses coming

How it works ↓

  1. $5

    For your friend at signup

    Reserved after verified waitlist signup. No card or API key needed.

  2. +$10

    On their first Pro or Max payment

    Planned bonus brings their total to $15, separate from monthly credit.

  3. $15

    For you when they subscribe

    Share your personal link. Earn credit after their first qualifying paid Pro or Max payment.

Reserve signup credit now. Paid subscription and $15 referral rewards are not active yet; signup alone earns no paid bonus. Inference remains closed.

Signup credit is limited to the existing funded campaign’s first 1,000 eligible accounts. Confirmed email is required; one signup reward per person. Existing $10 grants are preserved. The friend’s $15 is $5 + $10, not another $15 on top. Self-referrals, duplicate accounts, trials, failed payments and renewals would not qualify. Paid bonuses are coming with subscriptions; final eligibility, refund and campaign terms will be available before purchase. Promotional credit has no cash value and does not reserve compute.

Join the waitlist

Get started

Works with your stack.

Use the coding agents and SDKs you already have. Copy one prompt, or point your SDK at Lyber. Your key stays in your environment.

SDK and HTTP: any client that accepts an OpenAI-compatible base URL. CLI: Lyber becomes Codex's model. Prompt: the agent wires Lyber into your project; its own model stays the same.

Set it up.

  1. Get your API key.

    Request access with a verified email. Once approved, create a key in your dashboard and set it as LYBER_API_KEY.

    Request access →
    export LYBER_API_KEY=lyber_prepaid_…

  2. Copy the setup prompt.

    Set up Lyber in this project.

Lyber is an OpenAI-compatible API for uncensored, abliterated models.
- Base URL: https://lyber-paid-acceptance-gateway-production.up.railway.app/v1
- API key: read it from the LYBER_API_KEY environment variable. Never write the key into code, config files, logs or git.
- Model: lyber/auto. Lyber picks a ready model and the response's "model" field names the one that answered. List exact model IDs with GET {base URL}/models.

Do this:
1. Find where this project calls an LLM, or add a small client if it has none. Use the official OpenAI SDK for the project's language with its base URL set to the Lyber base URL.
2. Add LYBER_BASE_URL and LYBER_API_KEY to .env.example without values, and load them the way this project loads its other settings.
3. Make one test request with model "lyber/auto" and print the "model" field of the response.
4. Tell me what you changed. Don't commit, and don't print the key.

  3. Paste it into your agent.

    It edits your project and runs one test request. Review the changes before you commit.

    Works in Claude Code Codex Cursor+ any coding agent

Example session · Setup prompt
  1. youSet up Lyber in this project. …
  2. readpackage.json · src/lib/ai.ts
  3. editsrc/lib/ai.ts baseURL ← LYBER_BASE_URL
  4. edit.env.example + LYBER_BASE_URL + LYBER_API_KEY
  5. runnode scripts/lyber-check.mjs
  6. model: lyber/qwen38-27b-abliterated
  7. Lyber is wired in. 2 files changed; the key stays in your environment.

Illustrative output. Inference opens after each model is qualified.

Models & pricing

Heavy models. Open answers.

Five models from four leading families, with modified uncensored or abliterated weights only.

Preparing launch

Launch target · USD per 1M tokens · input / output

Qwen 3.8 27B Abliterated

1× A100 80GB

lyber/qwen38-27b-abliterated

$0.30 / $1.50

Cached input: $0.03 per 1M tokens

DeepSeek V4.1 Flash Uncensored

mixed FP8/FP4

lyber/deepseek41-flash-abliterated

$0.30 / $1.20

Cached input: $0.0075 per 1M tokens

GLM 5.3 · Full Uncensored

8× H200 FP8 candidate

lyber/glm53-full-fp8-abliterated

$1.50 / $4.00

Cached input: $0.25 per 1M tokens

GLM 5.3 · Flash Uncensored

4× H200 candidate

lyber/glm53-flash-abliterated

$0.14 / $0.45

Cached input: $0.025 per 1M tokens

Kimi K3 Abliterated

8× B300

lyber/kimi-k3-abliterated

$3.25 / $15.00

Cached input: $0.30 per 1M tokens

Inference is unavailable until each model and its hardware are qualified. Launch targets are provisional; the API's published price version governs billing.

Full models & pricing →

New at Lyber

Built with the model creators.

We work directly with dealign.ai’s model-building team and pay them for model development. Follow the exact checkpoints we’re qualifying for Lyber.

Uncensored derivative

Lyber’s MiniMax M3.1 Flash

In qualification · inference unavailable

Our uncensored derivative of MiniMaxAI’s MiniMax M3.1 Flash, published under lyber-ai. Runtime, tools, cache behavior and completed-task quality are being qualified. Hugging Face weight access is gated separately.

lyber-ai/MiniMax-M3.1-Flash-UNCENSORED-NVFP4

Pricing will follow serving qualification.

View checkpoint — Lyber’s MiniMax M3.1 Flash ↗

Uncensored checkpoint

dealign.ai DeepSeek V4.1 Flash 1017

In qualification · inference unavailable

We’re qualifying dealign.ai’s uncensored 1017 checkpoint for long-context agent work. The checkpoint and its revision stay explicit; a base model or an older uncensored release is a different offering.

dealignai/DeepSeek-V4.1-Flash-1017-UNCENSORED-FP8

Provisional token targets appear in Models & pricing.

View checkpoint — dealign.ai DeepSeek V4.1 Flash 1017 ↗

Use cases

Built for work other models refuse.

Lyber answers, and you stay responsible for how it is used.

AI safety & platform teams

Read more: AI red teaming 

POST /v1/chat/completions
{
  "model": "lyber/auto",
  "messages": [
    { "role": "system",
      "content": "<your policy>" },
    { "role": "user",
      "content": "<your brief>" }
  ]
}

Your policy, not oursLyber adds no refusals of its own.

Request access →

Early access

Bring your intent.

Join the access list with a verified email. We review each request before keys and models open, and invites move you up the queue.

Get startedConnect Codex