KENOS is now in Pre-Production

< Get the App >

Launch KENOS
Launch KENOS

Future of KAI

Kohenoor Technologies' Roadmap for Accessible, Private and Intelligent AI

Founder, Kohenoor Technologies

10/11/20268 min read

From lightweight AI assistants running on everyday devices to sophisticated, cloud-powered intelligence, Kohenoor Technologies is building an AI ecosystem designed for everyone.

Artificial intelligence is advancing rapidly, but one fundamental challenge remains: how do we make powerful, practical AI accessible to individuals, professionals and businesses without requiring expensive hardware, complicated infrastructure or extensive technical knowledge?

At Kohenoor Technologies, we believe the future of artificial intelligence should not be limited to massive data centers, costly subscriptions or high-performance computing systems.

AI should be accessible, practical, affordable and, wherever possible, capable of running directly on the devices people already own.

This philosophy drives the development of the KAI family of intelligent models, beginning with KAIPAL Lite and progressively expanding toward more capable assistants, multimodal intelligence, advanced workflows and the full KAI ecosystem.

Our roadmap represents a carefully structured journey from miniature, locally running AI models to sophisticated cloud-powered intelligence integrated with the Kohenoor Operating System (KENOS).

And the journey has already begun.

One AI Ecosystem, Different Levels of Intelligence

Not everyone needs the same level of artificial intelligence.

A student exploring career opportunities, a shopkeeper managing inventory, an entrepreneur evaluating a business idea and an enterprise coordinating complex operations have very different requirements.

Providing the same heavyweight AI model to every user is neither necessary nor computationally efficient.

Instead, Kohenoor Technologies is developing a progressive family of AI models, each designed around specific capabilities, user needs and hardware requirements.

AI ModelVersionsCore CapabilitiesHardware RequirementsAccessKAIPAL Lite1, 1.1, 1.2Everyday business, finance and personal development advisoryNo dedicated GPU requiredFreeKAIPA1.13, 1.14, 1.15Expanded intelligence, image understanding and audio input8–12 GB GPUFreeKAIS (KAI Spirit)1.16–1.19Broader advisory capabilities and lightweight multi-step workflows12–24 GB GPUPaidSuper KAI2.0 onwardFull KAI intelligence, advanced tools and KENOS workflowsCloud-powered, no local hardware requiredPaid

This structure follows a straightforward principle: make essential AI capabilities freely accessible and introduce more sophisticated intelligence as users' needs grow.

KAIPAL Lite and KAIPA are planned as free offerings, while KAIS and Super KAI form the premium segment of the ecosystem.

KAIPAL Lite: Bringing AI to Everyday Computers and Mobile Devices

The first milestone in our roadmap is already available.

KAIPAL Lite (KAI Pocket Assistant Lite) represents our initial step toward practical, locally running artificial intelligence for mass adoption.

Unlike large language models that depend heavily on cloud infrastructure, KAIPAL Lite is designed to operate on modest computing hardware without requiring a dedicated graphics card.

The initial release features approximately 2 billion active parameters and a memory footprint of roughly 4 GB, making it suitable for a wide range of devices.

Its focus is not to compete with massive frontier models across every possible benchmark. Instead, its purpose is to provide meaningful utility in areas where people need assistance most frequently.

These include:

  • Business planning, pricing and customer management.

  • Everyday financial understanding and structured calculations.

  • Entrepreneurship and small-business decision support.

  • Professional development and career guidance.

  • Business communication and practical writing.

  • General advisory support for routine activities.

KAIPAL Lite is designed to operate offline, keeping user interactions on the local device when used without external services.

This approach offers an important advantage: greater control over personal information, reduced dependence on internet connectivity and access to AI without recurring cloud inference charges.

The KAIPAL Lite Evolution

The initial KAIPAL Lite roadmap includes three releases.

KAIPAL Lite 1: The foundation model, designed for smaller computers and compatible mobile devices. Available now.

KAIPAL Lite 1.1: An enhanced version targeting premium mobile devices and standard business computers. In preparation.

KAIPAL Lite 1.2: A further development intended for modern mid-range computers, with improvements guided by practical testing. Planned.

Our approach emphasizes real-world usefulness rather than parameter counts alone.

Every release must demonstrate measurable improvements and pass internal testing before publication.

KAIPA: Expanding AI Beyond Text

The next major stage introduces KAIPA (KAI Pocket Assistant).

Where KAIPAL Lite concentrates on lightweight text-based intelligence, KAIPA is planned to expand the interaction between users and AI by incorporating visual and auditory understanding.

Imagine a small-business owner photographing an invoice and asking AI to explain its contents, or a professional providing an audio recording instead of typing a lengthy question.

These are the kinds of practical interactions KAIPA is being designed to support.

Its planned capabilities include image interpretation, short audio input, deeper contextual understanding and expanded advisory functions.

KAIPA is expected to require a graphics card with approximately 8–12 GB of graphics memory.

Despite these additional capabilities, KAIPA is planned to remain free, continuing our commitment to accessible, locally running artificial intelligence.

KAIS: Bringing the Spirit of Full KAI to Local Computing

As AI tasks become more sophisticated, the challenge shifts from answering individual questions to understanding broader objectives and completing connected activities.

This is where KAIS (KAI Spirit) enters the roadmap.

KAIS is envisioned as a more comprehensive local AI model that brings together a larger portion of KAI's advisory capabilities.

Rather than focusing exclusively on a single response, KAIS is planned to support lightweight multi-step workflows.

For example, an entrepreneur may want AI to examine a business situation, identify operational weaknesses, compare possible improvements and organize recommendations into a practical action plan.

Such tasks require more depth, consistency and contextual reasoning than typical lightweight interactions.

KAIS is being designed to address this intermediate level of intelligence.

The proposed hardware requirement is a GPU with approximately 12–24 GB of graphics memory.

KAIS will be the first paid model within the locally running KAI family. Its distribution arrangements and commercial details will be announced closer to release.

Super KAI: The Full Intelligence Experience

At the advanced end of the roadmap stands Super KAI, beginning with version 2.0.

Super KAI represents the planned evolution toward full KAI capabilities, bringing together advanced advisory intelligence, integrated tools, coordinated workflows and the broader KENOS environment.

Unlike the locally running pocket models, Super KAI is designed to operate through Kohenoor Technologies' cloud services.

This removes the need for users to own expensive computing hardware while allowing access to substantially greater computational resources.

Future releases, beginning with version 2.1, are intended to expand its abilities through additional skills, tools and integrations.

Super KAI is also designed around an essential principle: human supervision must remain central to consequential decisions.

AI can analyze, advise, organize and assist with execution, but human judgment and authorization remain fundamental to the system.

KENOS: More Than a Standalone AI Assistant

The KAI roadmap is not simply a sequence of language models.

It forms part of a broader technological vision centered on the Kohenoor Operating System (KENOS).

KENOS is designed as an integrated, chat-native operating environment where users can interact with applications, services and intelligent capabilities through a unified interface.

Within this ecosystem, KAI serves as the intelligence layer connecting users with practical applications and workflows.

The long-term objective is to reduce the complexity of digital interactions by allowing individuals and organizations to communicate their requirements naturally rather than navigating numerous disconnected tools.

The relationship between KAI and KENOS is therefore central to the roadmap.

As KAI evolves, its ability to support the KENOS environment is expected to grow alongside it.

Local AI and Cloud AI: Two Approaches, One Vision

One of the most important distinctions within the KAI ecosystem is the difference between locally running models and cloud-based intelligence.

Local AI prioritizes accessibility, privacy and independence.

It allows users to run supported models on their own machines, potentially without internet connectivity and without sending routine interactions to external inference services.

Cloud AI prioritizes computational capability, advanced integrations and sophisticated workflows.

It enables access to larger models and services that may be impractical to run on ordinary personal devices.

We do not view these approaches as competitors.

They serve different purposes, and both have an important role in the future of artificial intelligence.

Our objective is to provide users with appropriate choices according to their needs, hardware resources and desired capabilities.

The Principles Behind Every KAI Model

While the capabilities of individual models will expand, the fundamental philosophy of KAI remains consistent.

Human-centered intelligence: KAI is built to support human decisions rather than replace human judgment.

Practical utility: Models are designed around meaningful, everyday applications rather than benchmark performance alone.

Responsible advisory behavior: KAI is intended to explain options, risks and trade-offs without presenting uncertain predictions as facts or making consequential decisions on behalf of users.

Transparency about limitations: Models should acknowledge uncertainty, avoid invented information and communicate the boundaries of their capabilities.

Language accessibility: The KAI family is intended to serve users across languages and backgrounds.

Privacy and user control: Locally running models provide an option for users who prefer to keep their interactions on their own devices.

Human supervision: Advanced workflows must preserve appropriate human oversight and authorization.

These principles guide the development of KAIPAL Lite, KAIPA, KAIS and Super KAI.

KAI Is Already Accessible Worldwide

Users do not need to wait for the completion of the roadmap to begin exploring KAI.

KAI is already accessible through the Kohenoor online platform.

Access KAI online: https://www.kohenoor.net

The service is available through a web browser without requiring users to install the locally running models.

KAI and Super KAI are proprietary products offered through Kohenoor's services. Direct access is provided through subscription arrangements, while API access for developers and businesses is planned for the future.

Dedicated web, Android and iOS applications are also included in the development direction, with the aim of providing a consistent experience across devices.

Download KAIPAL Lite

For users who prefer locally running artificial intelligence, KAIPAL Lite is available through the following platforms:

Hugging Face: https://huggingface.co/KOHENOOR-AI/kai-lite1

Ollama: https://ollama.com/kohenoor/kai-lite1

Future pocket-model releases will be announced through Kohenoor's official channels as they become available.

Development Without Artificial Deadlines

The KAI roadmap is deliberately capability-driven rather than deadline-driven.

Artificial intelligence development is an iterative process involving architecture decisions, model training, evaluation, optimization, practical testing and continuous improvement.

A larger model is not automatically a better model, and additional features do not necessarily translate into greater usefulness.

Each release must justify its place in the ecosystem through demonstrated capabilities.

For this reason, the roadmap does not establish fixed release dates.

Planned model specifications, features and hardware requirements may evolve as development progresses.

Our commitment is to release models when they are ready, not merely when a calendar demands it.

A Future Where AI Belongs to Everyone

The next chapter of artificial intelligence will not be defined exclusively by the largest models or the most powerful data centers.

It will also be shaped by how effectively AI reaches ordinary people, integrates into daily activities and creates practical value.

A capable local assistant on a modest computer can be transformative for someone who does not have access to sophisticated infrastructure.

A multimodal assistant can simplify how professionals interact with information.

An advanced, cloud-powered intelligence platform can enable organizations to manage increasingly complex activities.

These are different levels of the same opportunity.

Through KAIPAL Lite, KAIPA, KAIS, Super KAI and KENOS, Kohenoor Technologies is working toward an AI ecosystem that grows with the user, from everyday assistance to advanced intelligent workflows.

Our vision is straightforward:

Powerful AI where it is needed. Accessible AI where it matters. Human judgment where it counts.

The roadmap is evolving. The foundation is already available.

And this is only the beginning.

About Kohenoor Technologies

Kohenoor Technologies develops AI-powered platforms, intelligent advisory systems and integrated digital solutions through its growing KAI and KENOS ecosystem.

Explore our platforms and follow future announcements:

Disclaimer: This roadmap reflects Kohenoor Technologies' current development direction. Proposed capabilities, specifications, availability and commercial arrangements are subject to modification based on research, development and testing. Future releases and features are not guaranteed.

KAI AI Roadmap, Kohenoor Technologies, KAIPAL Lite, KAIPA, KAIS, Super KAI, KENOS, Local AI, Offline AI Assistant, Privacy-First AI, Small Language Models, AI Without GPU, Free AI Models, Ollama AI, Hugging Face Models, On-Device AI, Edge AI, Multimodal AI, AI for Business, AI for Entrepreneurs, AI Model Development, Cloud AI, Agentic AI, Human-Centered AI, Responsible Artificial Intelligence