← Back to the wire

Train AI models with Unsloth and Hugging Face Jobs for FREE

AnnouncementModelFeb 20, 2026

Hugging Face and Unsloth announced free credits for fine-tuning models on Hugging Face Jobs, enabling users to train LiquidAI's LFM2.5-1.2B-Instruct through coding agents like Claude Code and Codex. Unsloth provides approximately 2x faster training and 60% less VRAM usage compared to standard methods. The 1.2B parameter model runs under 1GB of memory and is optimized for on-device deployment on CPUs, phones, and laptops.

Evidence

1source· awaiting independent confirmation

No score is assigned. Sources and their independence are shown in the citation chain below.

Citation chain · 1 source

01highPRIMARY
CodexModelClaude CodeModelUnslothCompanyben burtenshawPersonshaun smithPersonLiquidAICompanyDaniel (Unsloth)PersonMichael HanPersonMaxime LabonnePersonDaniel van StrienPersonLFM2.5-1.2B-InstructModelHugging FaceCompany
Canonical: https://huggingface.co/blog/unsloth-jobs