Lailaba AI

Lailaba AI is a compact language model developed by Abdullahi Ibrahim Lailaba (a student/researcher connected to the Federal University Birnin Kebbi). It is based on the Qwen3-0.6B architecture and customized for programming, technology, education, research, cybersecurity learning, and general technical assistance.

This repository provides the complete model checkpoint, including its weights and configuration, for use with compatible Hugging Face Transformers tooling. It is distributed as a standalone model rather than as a LoRA adapter.

Intended uses

Lailaba AI is intended for experimentation and assistance with tasks such as:

  • Programming, Python, and software development
  • Web development and APIs
  • Artificial intelligence and machine learning
  • Linux and command-line tools
  • Cloud computing and software integration
  • Cybersecurity education and defensive security concepts
  • Learning, research, and technical explanations

The model may also be used as a starting point for chat applications, educational tools, developer utilities, and other text-based applications. It can be further fine-tuned for specialized tasks.

Quick start

Install dependencies

Qwen3 support requires a compatible version of Transformers. Install or upgrade the required packages:

pip install -U "transformers>=4.51.0" torch accelerate

Load and run the model

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "lailaba/lailaba-ai"

tokenizer = AutoTokenizer.from_pretrained(model_id)

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto"
)

prompt = "Explain what an API is in simple terms."

inputs = tokenizer(
    prompt,
    return_tensors="pt"
).to(model.device)

with torch.no_grad():
    outputs = model.generate(
        **inputs,
        max_new_tokens=256,
        do_sample=True,
        temperature=0.7,
        top_p=0.9
    )

# Decode only the newly generated tokens.
new_tokens = outputs[0][inputs["input_ids"].shape[-1]:]
response = tokenizer.decode(
    new_tokens,
    skip_special_tokens=True
)

print(response)

Conversational use

For chat applications, use the tokenizer's built-in chat template when it is available. The following example formats a system message and a user message before generating a reply.

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "lailaba/lailaba-ai"

tokenizer = AutoTokenizer.from_pretrained(model_id)

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype="auto",
    device_map="auto"
)

messages = [
    {
        "role": "system",
        "content": "You are Lailaba AI, a helpful technical assistant."
    },
    {
        "role": "user",
        "content": "What is Python?"
    }
]

inputs = tokenizer.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    return_tensors="pt"
).to(model.device)

with torch.no_grad():
    outputs = model.generate(
        inputs,
        max_new_tokens=256
    )

new_tokens = outputs[0][inputs.shape[-1]:]
response = tokenizer.decode(
    new_tokens,
    skip_special_tokens=True
)

print(response)

The chat example assumes that the tokenizer configuration in this repository includes a compatible chat template.

Example

User: What is an API?

Lailaba AI: An API (Application Programming Interface) is a set of rules that allows different software applications to communicate. One application can send a request to an API, and the API can return information or perform an operation for it.

This is an illustrative example, not a benchmark result or a guarantee of model behavior.

Cybersecurity use

The model can be used to explore cybersecurity concepts, secure programming, Linux security, network-security fundamentals, and defensive security practices.

Users are responsible for ensuring that security-related activities are authorized and comply with applicable laws and organizational policies.

Limitations and responsible use

Lailaba AI is a compact experimental model. Its size may limit its ability to handle complex reasoning, long instructions, advanced programming tasks, and specialized technical questions.

The model may:

  • Generate incorrect or incomplete information.
  • Produce code with bugs or security weaknesses.
  • Misinterpret prompts or follow instructions inconsistently.
  • Generate plausible-sounding but unsupported claims.
  • Lack up-to-date information about recent events and technologies.

Do not rely on its outputs as the sole basis for important decisions. Review, test, and validate generated content---especially code and security-related guidance---before using it.

Evaluation

Formal benchmark results have not yet been published for this release. No benchmark scores or performance claims are made in this model card. Future evaluations may cover programming, instruction following, technical question answering, mathematics, and general reasoning.

Model lineage and development

Lailaba AI is based on the Qwen3-0.6B architecture and is part of the Lailaba AI project. It is intended to evolve through further experimentation, evaluation, and improvements to training and deployment.

For details about the upstream architecture and its terms, consult the Qwen3 model documentation.

Repository contents

The repository contains the files needed to load the standalone checkpoint. Depending on the export, these may include:

lailaba-ai/
β”œβ”€β”€ config.json
β”œβ”€β”€ generation_config.json
β”œβ”€β”€ tokenizer_config.json
β”œβ”€β”€ tokenizer.json
β”œβ”€β”€ special_tokens_map.json
β”œβ”€β”€ model.safetensors
└── README.md

The actual filenames and number of weight shards may differ in the published repository.

License

This repository is marked as Apache-2.0. The model is based on Qwen3-0.6B; review the upstream model repository and applicable license terms when redistributing or using the model commercially.

Developer

Abdullahi Ibrahim Lailaba

Feedback and contributions

Feedback, testing, and independent evaluations are welcome. Please report reproducible issues with the prompt, expected behavior, and actual output where possible.


Lailaba AI --- Build. Learn. Create.

A compact experimental language model for programming, technology, education, research, and technical assistance.

Downloads last month
813
Safetensors
Model size
0.6B params
Tensor type
F32
Β·
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support