Thomson Reuters Launches Its Own Frontier Model (www.thomsonreuters.com)

Thomson Reuters has launched its first proprietary large language model, Thomson, developed in-house at a fraction of the cost typically associated with frontier models. Built on a robust open-source foundation, Thomson has been fine-tuned using decades of specialized content from authoritative sources like Westlaw and Reuters, enabling it to execute complex instructions and navigate dense, domain-specific knowledge with high accuracy. This model promises a shift in the AI landscape, emphasizing efficiency and cost-effectiveness over sheer scale, positioning Thomson Reuters uniquely with complete control over both the model and its data. The significance of Thomson lies in its commitment to AI sovereignty, addressing critical concerns around how models are trained and the privacy of user data. Thomson's early evaluations indicate that it performs comparably to leading frontier models while providing enhancements that arise from its proprietary training and expert-driven design process. As it integrates into Thomson Reuters' legal tools, particularly CoCounsel Legal, this launch marks a pivotal evolution in professional AI, setting a new standard for trust and accountability in AI applications, particularly within high-stakes environments like law and finance.