KILL LABS
WIPJoin early access

Independent AI lab · In development

Kill constraints.
Keep control.

Open-model infrastructure for people who need more control over their models, data and deployment.

Join early access Explore the platform
Open models
Custom fine-tuning
Private inference
Developer APIs

The open-model ecosystem

Great models. A KILL future.

The models you know.
More control, coming next.

  • Llama

    Meta
    KILL LabsComing soon

  • Qwen

    Alibaba
    KILL LabsComing soon

  • DeepSeek

    DeepSeek
    KILL LabsComing soon

  • Mistral

    Mistral AI
    KILL LabsComing soon

  • Gemma

    Google DeepMind
    KILL LabsComing soon

  • gpt-oss

    OpenAI
    KILL LabsComing soon

  • Kimi

    Moonshot AI
    KILL LabsComing soon

  • GLM

    Z.ai
    KILL LabsComing soon

  • MiniMax

    MiniMax
    KILL LabsComing soon

  • Phi

    Microsoft
    KILL LabsComing soon

  • Nemotron

    NVIDIA
    KILL LabsComing soon

  • Granite

    IBM
    KILL LabsComing soon

Planned support · Versions and availability to be confirmed.

01 / The mission

Your model should
work for you.

Access to the weights is only the beginning. Understand what changed, choose what remains, and control where your model runs.

KILL LABS / COMPARE Animated preview
Your prompt

Default model

Standard

A generic restriction. No context considered.

Conversation stopped

KILL Labs

Operator control

Context considered. Response delivered.

Illustrative scenarios, not live model outputs or benchmark results.

02 / The platform

One stack. Your rules.

In development
01

Open-model liberation

Inspect model behavior. Reduce unnecessary constraints.

Research into abliteration and related techniques for more deliberate control of open models.

02

Custom fine-tuning

Adapt models to your domain, tone and workflow.

Purposeful model adaptation, shaped around your operational requirements.

03

Private inference

Keep sensitive workloads under your control.

Infrastructure being designed around your prompts, your data and your deployment choices.

04

Developer APIs

Bring selected models into the tools you already use.

Developer-friendly access to the models and capabilities we are building.

03 / Our approach

Freedom without guesswork.

  1. 00

    Open model

    Start with inspectable open weights.

  2. 01

    Analyze

    Understand behavior and its limitations.

  3. 02

    Transform

    Apply deliberate, documented changes.

  4. 03

    Evaluate

    Measure capabilities and regressions.

  5. 04

    Deploy

    Run privately. Control how you give access.

04 / Principles

Control requires
responsibility.

Open by foundation

Models whose weights and behavior can be inspected.

Private by design

Sensitive workloads should stay under the operator’s control.

Measurable changes

Modifications should be evaluated, documented and reproducible.

Built carefully.
Released when it’s real.

Research

Active

Infrastructure

In development

Early access

Preparing

05 / Questions

A few things,
answered.

An independent AI lab focused on open models, model customization, private inference and developer access.

Ready to take back
control?

Join the early-access list. Follow what comes next.

Development updates. No spam.