Step 1

Upload Training Data

Import your JSONL dataset to begin fine-tuning

Drop your JSONL file here or browse files

Format Requirements

✓ JSONL format with messages array
✓ Minimum 50 examples (100-1000 recommended)
✓ Each message: role and content
✓ Valid roles: system, user, assistant

Example Format

{"messages": [
  {"role": "user", "content": "..."},
  {"role": "assistant", "content": "..."}
]}

1/2

2/2

Step 3

Model Recommendation

Based on your dataset and requirements

Analyzing your dataset...

What Happens Next

📓

One-Click Colab Notebook

TuneKit will generate a ready-to-run Google Colab notebook with:

✨

Everything pre-configured: Your dataset embedded, model selected, LoRA hyperparameters optimized, and all training code ready. Just click "Run All" in Colab and your model will be fine-tuned in ~15 minutes on a free T4 GPU.

⚡

Fast & Free Training

Training powered by Unsloth for maximum efficiency:

🚀 2x Faster

Unsloth optimization

💾 70% Less VRAM

Runs on free GPUs

💰 $0 Cost

Google Colab free tier

No infrastructure, no credit card, no setup required.

Step 4

Train Your Model

Choose how you want to train your model

Google Colab

Train with free T4 GPU. No setup required.

Free T4 GPU (15GB VRAM)
Dataset embedded in notebook
Just click "Run All"

Download Package

Get training scripts for your own hardware.

Full control over training
Use your own GPU
Customize as needed

Upload Training Data

Model Recommendation

-

Alternative Models

What Happens Next

One-Click Colab Notebook

Fast & Free Training

Train Your Model

Google Colab

Download Package

Notebook Downloaded!

Package Ready!

Upload Training Data

Add System Prompt

Model Recommendation

-

Alternative Models

-

Why choose this model?

What Happens Next

One-Click Colab Notebook

Fast & Free Training

Train Your Model

Google Colab

Download Package

Notebook Downloaded!

Package Ready!