Train AI models with Unsloth and Hugging Face Jobs for FREE

2026-02-20 · Source: Hugging Face - Blog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Software Development & Engineering · Depth: Intermediate, short

Summary

Hugging Face, in collaboration with Unsloth, now offers a streamlined method for fine-tuning large language models (LLMs) like LiquidAI/LFM2.5-1.2B-Instruct using Hugging Face Jobs, with free credits available through the Unsloth Jobs Explorers program. Unsloth accelerates training by approximately 2x and reduces VRAM usage by about 60% compared to standard approaches, making small model fine-tuning highly cost-effective, potentially costing only a few dollars. This process can be automated using coding agents such as Claude Code or Codex, which leverage Hugging Face skills to generate and submit training scripts to managed cloud GPUs. The fine-tuned models, optimized for on-device deployment, can run on CPUs, phones, and laptops, enhancing accessibility and practical application.

Key takeaway

For AI Engineers seeking to fine-tune LLMs efficiently and affordably, leveraging Unsloth with Hugging Face Jobs is a compelling option. You can significantly reduce training costs and VRAM consumption, even automating the process with coding agents. Consider joining the Unsloth Jobs Explorers to access free credits and experiment with fine-tuning models like LiquidAI/LFM2.5-1.2B-Instruct for diverse deployment scenarios, including on-device applications.

Key insights

Unsloth and Hugging Face Jobs enable efficient, cost-effective LLM fine-tuning, automatable via coding agents.

Principles

• Small LLMs are cost-effective for fine-tuning.
• Unsloth significantly reduces VRAM and training time.

Method

Install the `hf` CLI or Hugging Face skill in a coding agent (e.g., Claude Code, Codex), then prompt the agent to train a model on Hugging Face Jobs, specifying model, dataset, and Unsloth usage.

In practice

• Fine-tune LiquidAI/LFM2.5-1.2B-Instruct for on-device deployment.
• Use `hf jobs` CLI for direct job submission.
• Request Trackio monitoring for real-time loss curves.

Topics

• LLM Fine-tuning
• Unsloth
• Hugging Face Jobs
• Coding Agents
• Small Language Models

Code references

• huggingface/blog
• unslothai/unsloth
• huggingface/skills
• huggingface/skills

Best for: AI Engineer, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• unslothai / unsloth — Unsloth significantly accelerates LLM fine-tuning and reduces VRAM, enabling longer context windows and broader model support.
• Self-Training Agents: Hermes Agent, HF Traces, Skills, MCP & Finetuning — Merve Noyan, Hugging Face — Hugging Face provides a comprehensive open-source ecosystem for developing, deploying, and managing AI agents and models.
• Scale LLM fine-tuning with Hugging Face and Amazon SageMaker AI — Specialized LLMs fine-tuned on proprietary data offer enterprises accuracy, security, and domain-specific knowledge.
• You NEED to try these open-source AI projects RIGHT NOW — Open-source AI projects offer diverse solutions for enhanced search, local document processing, agentic engineering, and LLM cost reduction.
• YES: Harness Self-optimization w/ 9B LLM (Local AI) — Small LLMs can update agent harnesses effectively, but solver model capability dictates the benefit from these updates.
• Gemma 4 Fine-Tuning on Single GPU | Training Gemma 4 With Unsloth on Custom Dataset (🔴 Live) — Fine-tuning LLMs may offer better accuracy and lower GPU RAM than prompting alone.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Hugging Face - Blog.