Documents
Create, analyse, edit, and transform working materials.
ORCNEIT's own language model
ORCNEITGPT is a language model under development at ORCNEIT and the technological foundation for the company's future system of intelligent tools.
The project is being built as more than a model for generating text. ORCNEITGPT's long-term direction extends from a universal everyday assistant to a more advanced work environment for analysis, programming, documents, presentations, research tasks, and interaction with external services.
The project is currently in active development, including model training and experimental evaluation.
01Development direction
Universal assistant
One of ORCNEITGPT's foundational directions is an intelligent assistant for everyday work.
- 01Questions
- 02Complex topics
- 03Writing
- 04Ideas and planning
- 05Information analysis
- 06Preparing materials
- 07Learning and research
- 08Everyday knowledge work
This mode depends on clear interaction, stable dialogue, an ability to account for context, and predictable model behaviour.
We see the universal assistant as the foundational layer of the future ORCNEITGPT system, with more specialised modes built on top of it.
02Development direction
Assistant for complex work
As the model develops, ORCNEITGPT is intended to support not only short prompts, but also tasks that require deeper analysis and sustained work.
- 01
Gather
Large volumes of information, multiple sources, and documents.
- 02
Understand
Project context, alternatives, connections, and contradictions.
- 03
Plan
A complex route through interdependent stages of work.
- 04
Research
Materials and interim results held in a sustained context.
- 05
Form
A structured deliverable rather than a single short response.
- 06
Verify
Intermediate conclusions and the completed result.
The model should do more than return one answer: it should help move from the initial task to a completed result.
This mode depends not only on the language model itself, but also on memory, search, file handling, context management, and verification of intermediate results.
03Development direction
Agentic work
By an agent, we mean a system that can do more than respond: it can carry out a sequence of actions with the tools available to it.
Presentations
Prepare structure, writing, visual content, and later revisions.
Programming
Work with a codebase, make changes, find errors, test, and complete engineering tasks.
Data
Process tables, structured information, computations, and research materials.
Projects
Carry out several interdependent stages in sequence.
Agentic functions require substantially stricter control than an ordinary conversation with a model.
The ability to take external actions will be developed separately from foundational language capabilities and introduced only where the system can reliably understand the task, the boundary of its authority, and the result of an action.
04Development direction
Programming
In the future, the system should work not only with isolated code fragments, but with a broader engineering context.
Project context
- 01Explain existing code
- 02Find errors and potential problems
- 03Propose changes
- 04Create new components
- 05Work across multiple files
- 06Assist with testing
- 07Analyse logs and execution results
- 08Account for project structure and the task
A more advanced agentic mode could potentially perform some of these actions directly in the user's working environment.
Access to projects, command execution, and file modification require a separate layer of security and control. These capabilities are not treated as a required part of the first ORCNEITGPT versions.
05Development direction
Documents and presentations
ORCNEITGPT is intended to evolve beyond returning text in a chat and towards working directly with professional materials.
- 01Prepare documents
- 02Analyse existing files
- 03Edit content
- 04Structure information
- 05Form reports
- 06Create tables
- 07Prepare presentations
- 08Revise existing presentations
- 09Work with content, structure, and visual representation
The objective is to reduce the distance between a user's request and a finished result.
Instead of producing text that a user must manually transfer elsewhere, the system should gradually move towards creating and changing the working material itself.
06Development direction
Integrations
Interaction with external services is one of ORCNEITGPT's long-term directions.
01Email
02
Calendars
03
Cloud storage
04
Document systems
05
Code repositories
06
Project management
07
Enterprise systems
08
Secure APIs
Integrations should allow ORCNEITGPT to work with the user's real context and perform permitted actions within the systems they already use.
Connecting an external service must not automatically give the model unrestricted access to it.
07Development direction
From answer to result
ORCNEITGPT's long-term direction is a gradual move from a simple exchange to a broader working sequence.
01Request
02Text response
- 01Task
- 02Context
- 03Planning
- 04Tool use
- 05Execution
- 06Verification
- 07Result
The language model remains the central component, while a software environment gradually forms around it to work with information and tools beyond an ordinary text conversation.
This is why ORCNEITGPT is being developed as both a model and a technology platform.
08Development direction
One system, different levels of work
We do not necessarily see the everyday assistant, professional tools, and an agent as separate products.
Assistant
An ordinary request without a complex agent environment.
Tools
Files, professional context, and specialised work.
Agent
Planning and several consecutive actions for a complex task.
In the long term, they may become different levels of one ORCNEITGPT system.
The system should select the level of complexity appropriate to the task while preserving one clear interface for the user.
09Development direction
Development will be gradual
The project must first establish a robust language foundation.
- 01Model maturity
- 02Software infrastructure
- 03Safety
- 04Extended validation
More advanced capabilities require a mature model, dedicated software infrastructure, safety mechanisms, and extended validation.
ORCNEIT does not claim that every direction described here will be available in the first public version, and does not set a date for their simultaneous release.
Individual functions may appear gradually, change during development, or go through limited testing before broader release.
The priority is not the greatest number of functions in the first release, but a foundation on which they can develop consistently and reliably.
Now
Current state
- Status
- In progress
- Stage
- Model training
At the current stage, the primary work is focused on training and developing the ORCNEITGPT language model itself.
In parallel, ORCNEIT is building the software and computing foundation needed for later stages of the project.
Public information about new capabilities will appear after they have actually been developed, evaluated, and are ready to demonstrate.
We prefer to describe working technology after achieving a result rather than turn development plans into promises of functions that have yet to be built.
Development direction
The direction of ORCNEITGPT
ORCNEITGPT is being built as a long-term project.
Today, its foundation is ORCNEIT's own language model.
Further development is intended to move towards a universal assistant, tools for complex intellectual work, agentic capabilities, and integrations with real working environments.
Each successive level must rest on capabilities that have been validated at the previous one.
First, the model.
Then, the tools.
Then, increasingly complex independent work on tasks.