📒Unsloth Notebooks

Train your own model with our notebooks, powered by free GPU compute. Click Run all (or save locally), add your dataset, train, and deploy. You can use any model in the notebooks.

GRPO (RL)Text-to-speechVisionUse-caseKaggle

Also see our GitHub repo for our notebooks: github.com/unslothai/notebooks

Colab notebooks

Standard notebooks:

• gpt-oss (20b) • Inference • Fine-tuning

• EmbeddingGemma (300M) - new

• Qwen3 (14B) • Qwen3-VL (8B)

• Qwen3-2507-4B • Thinking • Instruct

• Gemma 3 (4B) • Text • Vision • 270M • FunctionGemma

• Gemma 3n (E4B) • Text • Vision • Audio

• Mistral Ministral 3

• DeepSeek-OCR

• IBM Granite-4.0-H

• Phi-4 (14B)

• Llama 3.1 (8B) • Llama 3.2 (1B + 3B)

GRPO (Reasoning RL):

• gpt-oss-20b (automatic kernels creation)

• Mistral Ministral 3 (solving sodoku) - new

• Qwen3-8B - FP8 (L4) - new

• Llama-3.2-1B - FP8 (L4) - new

• gpt-oss-20b (auto win 2048 game)

• Qwen3-VL (8B) - Vision GSPO

• Qwen3 (4B) - Advanced GRPO LoRA

• Gemma 3 (4B) - Vision GSPO

• gpt-oss-20b (2048 OpenEnv example)

• DeepSeek-R1-0528-Qwen3 (8B) (for multilingual usecase)

• Gemma 3 (1B)

• Llama 3.2 (3B) - Advanced GRPO LoRA

• Llama 3.1 (8B)

• Phi-4 (14B)

• Mistral v0.3 (7B)

Text-to-Speech (TTS):

• Sesame-CSM (1B)

• Orpheus-TTS (3B)

• Whisper Large V3 - Speech-to-Text (STT)

• Llasa-TTS (1B)

• Spark-TTS (0.5B)

• Oute-TTS (1B)

Speech-to-Text (SST):

• Whisper-Large-V3

• Gemma 3n (E4B) - Audio

Vision (Multimodal):

• Qwen3-VL (8B)

• Mistral Ministral 3

• DeepSeek-OCR

• Paddle-OCR (1B)

• Gemma 3n (E4B)

• Gemma 3 (4B)

• Llama 3.2 Vision (11B)

• Qwen2.5-VL (7B)

• Pixtral (12B) 2409

• Qwen3-VL - Vision GSPO - new

• Qwen2.5-VL - Vision GSPO

• Gemma 3 (4B) - Vision GSPO

Embedding models:

• EmbeddingGemma (300M) - new

• Qwen3-Embedding 4B - new

• Qwen3-Embedding 0.6B - new

• BGE M3 - new

• ModernBERT-large - new

• All-MiniLM-L6-v2 - new

• GTE ModernBert - new

Large LLMs:

Notebooks for large models: These exceed Colab’s free 15 GB VRAM tier. With Colab’s new 80 GB GPUs, you can fine-tune 120B parameter models.

• gpt-oss-20b (500K context) - new

• GLM-4.7-Flash - new

• Nemotron-3-Nano-30B-A3B LoRA notebook - new

• NeMo Gym Sudoku GRPO notebook - new

• gpt-oss-120b

• Qwen3 (32B)

• Llama 3.3 (70B)

• Gemma 3 (27B)

• Baidu ERNIE 4.5 VL (28B) - new

Other important notebooks:

• Customer support agent

• Mistral Ministral 3 - new (solving sodoku)

• Deploy on LM Studio - new

• Quantization-Aware Training (QAT) - new

• Phone Deployment - new

• Reason before Tool Calling notebook - new

• Mobile Actions notebook - new

• Automatic Kernel Creation with RL

• ModernBERT-large - new Aug 19

• Synthetic Data Generation Llama 3.2 (3B)

• gpt-oss-20b (500K context) - new (A100)

• Tool Calling

• Mistral v0.3 Instruct (7B)

• Ollama

• ORPO

• Continued Pretraining

• DPO Zephyr

• Inference only

• Llama 3 (8B)

Specific use-case notebooks:

• Phone Deployment - new

• Deploy on LM Studio - new

• Reason before Tool Calling - new

• Mobile Actions - new

• Customer support agent

• Quantization-Aware Training (QAT) - new

• Automatic Kernel Creation with RL - new

• DPO Zephyr

• BERT - Text Classification - (AutoModelForSequenceClassification)

• Ollama

• Tool Calling

• Continued Pretraining (CPT)

• Multiple Datasets by Flail

• KTO by Jeffrey

• Inference chat UI

• Conversational

• ChatML

• Text Completion

Rest of notebooks:

• Qwen2.5 (3B)

• Gemma 2 (9B)

• Mistral NeMo (12B)

• Phi-3.5 (mini)

• Phi-3 (medium)

• Gemma 2 (2B)

• Qwen 2.5 Coder (14B)

• Mistral Small (22B)

• TinyLlama

• CodeGemma (7B)

• Mistral v0.3 (7B)

• Qwen2 (7B)

Kaggle notebooks

Standard notebooks:

• gpt-oss (20B) - new

• Gemma 3n (E4B)

• Qwen3 (14B)

• Magistral-2509 (24B) - new

• Gemma 3 (4B)

• Phi-4 (14B)

• Llama 3.1 (8B)

• Llama 3.2 (1B + 3B)

• Qwen 2.5 (7B)

GRPO (Reasoning) notebooks:

• Qwen2.5-VL - Vision GRPO - new

• Qwen3 (4B)

• Gemma 3 (1B)

• Llama 3.1 (8B)

• Phi-4 (14B)

• Qwen 2.5 (3B)

Text-to-Speech (TTS) notebooks:

• Sesame-CSM (1B)

• Orpheus-TTS (3B)

• Whisper Large V3 – Speech-to-Text

• Llasa-TTS (1B)

• Spark-TTS (0.5B)

• Oute-TTS (1B)

Vision (Multimodal) notebooks:

• Llama 3.2 Vision (11B)

• Qwen 2.5-VL (7B)

• Pixtral (12B) 2409

Specific use-case notebooks:

• Tool Calling

• ORPO

• Continued Pretraining

• DPO Zephyr

• Inference only

• Ollama

• Text Completion

• CodeForces-cot (Reasoning)

• Unsloth Studio (chat UI)

Rest of notebooks:

• Gemma 2 (9B)

• Gemma 2 (2B)

• CodeGemma (7B)

• Mistral NeMo (12B)

• Mistral Small (22B)

• TinyLlama (1.1B)

To view a complete list of all our Kaggle notebooks, click here.