AI summary: Stability AI introduces FreeWilly1 and FreeWilly2, open access Large Language Models (LLMs) with exceptional reasoning abilities. Built on LLaMA foundation models, they were fine-tuned using a synthetically-generated dataset. The training was inspired by Microsoft’s methodology, but with different data sources. Despite using a smaller dataset, the models show impressive performance across various benchmarks. They excel in intricate reasoning, understanding linguistic subtleties, and answering complex questions. The models are research experiments, released for open research under a non-commercial license.
Read more at Stability AI…