Thomson Reuters Launches Its Own Frontier Model

Thomson Reuters launched Thomson, a proprietary large language model trained on less than 10% of its proprietary content and $40 million in investment, achieving performance on par with frontier models while remaining fully owned and controlled. The model leverages decades of Westlaw, Practical Law, Checkpoint, and Reuters content, enabling precise instruction following and dense domain-specific reasoning. Thomson’s approach emphasizes AI sovereignty, reducing inference costs and allowing integration with Thomson’s own legal tools.