Open-Weight Language Models

Overview

Open-weight language models make trained model weights available so people can run them locally, adapt them to new tasks or languages, and examine how they behave. Researchers build them with dense or mixture-of-experts architectures, then improve them through instruction tuning, preference training, retrieval, and quantization for efficient deployment. Evaluations cover reasoning, coding, multilingual and domain tasks, as well as bias, safety, reliability, and energy use; capable open models can compete with proprietary systems in some settings, but performance varies across languages, tasks, and hardware. Wider access supports local control and reproducibility, while also making misuse easier and leaving open how to provide robust safeguards, trustworthy evaluations, and model-lineage accountability.

Latest papers

CardsList

65 papers6 papers in the last four weeks, down 33% on the four weeks before. 0.0% of all new papers.

  1. Instruction-Tuning Open-Weight Language Models for BPMN Model Generation

    Dec 12, 2025Gökberk Çelikmasat, Atay Özgövde, Fatma Başak AydemirBusiness Process ModelingOpen-Weight Language Models

  2. Open Foundation Models in Healthcare: Challenges, Paradoxes, and Opportunities with GenAI Driven Personalized Prescription

    Feb 4, 2025Mahdi Alkaeed, Sofiat Abioye, Adnan Qayyum +6Medical LLMsOpen-Weight Language Models