Juniper by Cinqic

Local intelligence,
built around the user.

Juniper is a lightweight, local-first AI assistant in development, designed to help people learn, create, think clearly, and solve practical problems while keeping their data and choices under their control.

Explore Juniper How Juniper works

Juniper 1 20M · Approximately 20M parameters · In development

Cinqic is launching August 27, 2026.

JJuniper / development preview● Local-first

You

Help me make a clearer plan.

Juniper

Let’s start with the next useful step, then keep the choices visible.

Thinking with user control in mindconcept only

Ask Juniper something…

01 — Purpose

Built to be useful before being impressive.

Juniper is being developed as a compact assistant that prioritizes clear conversation, practical tools, local operation, and honest limitations. The goal is not to build the largest model possible. The goal is to build an assistant people can understand, modify, and rely on for everyday work.

Lightweight

Designed around efficient local operation and modest hardware.

Local-first

Runs on user-controlled hardware whenever practical.

Helpful

Built to support learning, creation, reasoning, and real tasks.

Modifiable

Open components should remain inspectable and adaptable.

Development target

Juniper 1 20M

A compact foundation for a larger ecosystem.

Juniper 1 20M is an approximately 20-million-parameter decoder-only language model being developed from scratch as a lightweight foundation for fast, efficient, and helpful conversation. It is an early foundation for Juniper, not the complete assistant or ecosystem.

Status: In development
Target size: Approximately 20M parameters
Architecture: Dense decoder-only transformer
Primary interface: Text conversation
Priority: Lightweight local operation
Planned license: Open source
Planned release: August 27, 2026

Language modelToolsLocal retrievalMemorySpecialized agentsUser control

Concept design showing the planned Juniper assistant interface — Early Juniper UI concept — a conceptual preview, not a released application; interface labels are illustrative.

A product concept

One clear place to work with Juniper.

Juniper is being designed as an assistant and coordinating agent: a direct, quietly supportive interface for learning, practical work, and clear thinking. The visual shown here is an early concept for that future experience.

Juniper 1 20M is the underlying language-model project in development; Juniper is the broader assistant, interface, and long-term ecosystem it is intended to support.

02 — Future

Long-term ecosystem vision

The long-term goal is for Juniper to coordinate specialized agents through one simple interface. Each agent would focus on a specific type of work while Juniper handles conversation, coordination, context, and user control.

Junipercoordination

03 — Philosophy

Independent AI

Independent AI is the design philosophy guiding Juniper’s development. It is built around clear principles: tell the truth, admit uncertainty, protect user control, respect privacy, explain decisions clearly, and prioritize useful outcomes over confident-sounding answers.

Truth over confidence

Juniper should acknowledge uncertainty instead of inventing certainty.

Clarity over confusion

Responses should be direct, understandable, and useful.

Privacy by default

User information should remain under user control whenever practical.

User control

The assistant should support the user without quietly taking ownership of their decisions or data.

Independent AI exists to make Juniper more honest, practical, and understandable, not more restrictive or performative.

These principles are intended to guide training data, evaluations, tool behavior, and future preference training.

A product policy

AI should remain a choice.

Cinqic software is designed to remain useful without AI. Juniper adds optional capabilities while users stay in control of how much artificial intelligence they use.

Built by Cinqic

Artificial Intelligence & Software

Cinqic is being built by Markus Gillyard to create helpful open-source software that supports learning, creativity, clear thinking, and practical problem-solving while respecting users’ privacy and ownership.