Texo
Texo is a minimalist, open-source LaTeX OCR model with only 20M parameters, designed for fast and lightweight formula recognition. It achieves state-of-the-art performance by distilling PPFormulaNet-S and fine-tuning on UniMERNet-1M, delivering accuracy comparable to much larger models across simple printed, complex printed, scanned, and handwritten equation benchmarks. Texo runs directly in the browser, making it accessible without installation. It can be trained on consumer-level GPUs, and the full training pipeline is open-source, organized as a tutorial for self-reproduction. The project addresses the need for a free, fast, and accurate LaTeX recognition tool for STEM and AI note-taking, combining deep learning advances with a personal, open-source contribution.