---

About TellyAI

TellyAI is an independent AI organization focused on experimenting with and developing artificial intelligence models, datasets, and tools.

Our work currently focuses on early-stage research and experimentation. We are building models, preparing datasets, testing training approaches, and learning from the results.

The projects published by TellyAI may be incomplete, experimental, or still under development.

TellyAI is currently focused on research, testing, experimentation, and improvement.

We do not present early experimental models as finished production systems.

---

Models

Current TellyAI Research

Telly Oasis 1 Unreleased Research Model Telly Oasis 1 is an experimental TellyAI language-model project currently under development. Telly Oasis 1 has not been officially released as a finished model. The project is being developed and tested across training, inference, datasets, and model behavior. Status: In development Release: Not released Purpose: Research and experimentation View Repository →

Pre-1R Not Fully Trained Pre-1R is an early experimental stage of the TellyAI model project. Pre-1R is not fully trained. Its current outputs can be inaccurate, incomplete, inconsistent, or low quality. This is expected because development and training are still ongoing. Status: Active development Training: Not fully trained Purpose: Research and testing View Repository →

---

Telly Oasis 1

Unreleased Research Project

Telly Oasis 1 is an early TellyAI language-model project.

It is currently being developed and tested rather than released as a finished model.

Current Focus

• Model development
• Training experiments
• Dataset development
• Inference testing
• Conversational behavior
• Evaluation
• Future model improvements

Status

Important

Telly Oasis 1 is not currently a released production model.

The project may change significantly as training, testing, datasets, and model development continue.

---

Pre-1R

Early Development Model

Pre-1R is an early experimental stage of the TellyAI model project.

It is currently being used for development, testing, and experimentation.

Current Focus

• Training experiments
• Dataset experiments
• Model behavior testing
• Reasoning experiments
• Instruction following
• Evaluation
• Iteration and improvement

Status

Important

Pre-1R is not fully trained and should not be considered a finished or reliable AI assistant.

It may:

• Produce incorrect answers
• Produce incomplete answers
• Misunderstand instructions
• Behave inconsistently
• Perform poorly on some tasks

These limitations are part of the current research and training stage.

---

Research & Development

TellyAI is currently working through an iterative development process:

The objective is to use early experiments to identify problems, improve training approaches, and build better future models.

---

Datasets

telly-oasis-1

TellyAI also develops datasets for model training, experimentation, and evaluation.

The telly-oasis-1 dataset is associated with the Telly Oasis 1 project.

Dataset: telly-oasis-1

The dataset can be used as part of TellyAI's ongoing experimentation and model-development workflow.

Dataset Uses

The dataset may support experiments involving:

• Language modeling
• Instruction following
• Conversational data
• AI training
• Evaluation
• Fine-tuning experiments
• General model development

The dataset may be updated, expanded, cleaned, or changed as development continues.

---

Custom Datasets

Need a Dataset for Your AI Project?

TellyAI can also create custom datasets for you.

We can help create datasets tailored to a specific AI project, model, task, or training pipeline.

Custom datasets can be designed around your requirements rather than using a one-size-fits-all dataset.

Custom Dataset Types

We can work with requirements involving:

• Instruction-following datasets
• Conversation datasets
• Question-and-answer datasets
• Reasoning datasets
• Coding datasets
• Classification datasets
• Fine-tuning datasets
• Evaluation datasets
• Topic-specific datasets
• Structured AI training data

Custom Formats

Datasets can be structured for different workflows and pipelines, including:

JSONL · JSON · CSV · TXT · Parquet

The exact schema can be customized around the target model and training setup.

Custom Dataset Process

1. Requirements

You describe what you are building and what kind of data you need.

2. Dataset Design

The structure and format are planned around your intended use.

3. Creation

The dataset is generated and organized according to the agreed requirements.

4. Review

The dataset can be checked, cleaned, and refined.

5. Delivery

The completed dataset is provided in the required format.

Important

Custom dataset projects vary depending on requirements, dataset size, topic, complexity, and available resources.

Appropriate permissions, licenses, and legal requirements must be respected when working with external or copyrighted material.

Custom Dataset Requests

Have a dataset requirement?

Contact TellyAI through the available project channels and describe:

What you are building · What data you need · Approximate size · Desired format · Intended use

---

What TellyAI Builds

Models Language models, reasoning experiments, and future AI systems.

Datasets Public and custom datasets for AI development.

Research Experiments in training, evaluation, and model development.

Tools Future AI tools, applications, infrastructure, and APIs.

---

Future Projects

TellyAI Models

Future model generations based on lessons learned from current research.

Planned

Public Datasets

More datasets for AI training and experimentation.

In development

TellyAI Spaces

Interactive demonstrations and AI applications.

Coming soon

TellyAI API

Future API access to TellyAI models and services.

Planned

AI Tools

Developer tools and infrastructure for working with AI.

Planned

---

Roadmap

---

Explore TellyAI

---

Responsible Research

TellyAI's current models are experimental.

Their outputs should not automatically be treated as accurate, reliable, or authoritative.

Because Telly Oasis 1 remains unreleased and Pre-1R is not fully trained, their behavior can change substantially during development.

Users should review the documentation, model cards, datasets, and licenses associated with each individual project before using them.

---

Disclaimer

Telly Oasis 1 and Pre-1R are research and development projects.

Telly Oasis 1 has not been officially released as a finished model.

Pre-1R is not fully trained.

Both projects are currently being tested and improved. Their outputs may be inaccurate, incomplete, inconsistent, or poor.

They should not be relied upon for important decisions or represented as production-ready AI systems.

---