Thomson Reuters Launches Proprietary LLM 'Thomson' to Power CoCounsel
Thomson Reuters launched its proprietary large language model 'Thomson' powering new CoCounsel features.
Why it matters: This advancement offers legal professionals more precise, efficient AI tools for document review and legal workflows, influencing AI adoption in law firms and corporate legal departments worldwide.
- Thomson Reuters invested $40M to build its proprietary LLM 'Thomson' using less than 10% of its content library.
- Thomson powers CoCounsel Legal's new Tabular Analysis feature, handling up to 10,000 documents and 100 questions per review.
- CoCounsel Legal serves over one million users in 107 countries, now enhanced with Thomson's capabilities.
- The LLM was trained on data from Westlaw, Practical Law, Checkpoint, and Reuters platforms starting from open-source foundations.
Thomson Reuters has taken a major step in legal AI by launching its own proprietary large language model, Thomson. This model was developed at a cost of $40 million, covering talent and compute expenses, and was built upon open-source foundations. The training utilized content from Thomson Reuters' platforms including Westlaw, Practical Law, Checkpoint, and Reuters, deploying less than 10% of its extensive content library so far.
Thomson immediately powers a breakthrough feature in CoCounsel Legal: Tabular Analysis. Designed for high-volume document reviews, it allows attorneys to analyze up to 10,000 documents and pose up to 100 questions, returning answers in a dynamic, filterable table format. This integration aims to move legal professionals from research to trusted work product in one seamless workflow.
CoCounsel Legal is already widely adopted, with over one million users spanning 107 countries and territories. According to Thomson Reuters President Raghu Ramanathan, the legal industry is looking to move beyond AI that merely generates answers to solutions that convert insight into action across firms of all sizes.
Joel Hron, CTO of Thomson Reuters, emphasized building intelligence that is "highly capable, far more efficient and entirely under your control," while legal experts like Jennifer Eng from Polsinelli highlight CoCounsel’s ability to produce real work product rather than just responses.
This development marks a strategic effort by Thomson Reuters to deepen its AI integration and maintain leadership in legal technology by leveraging proprietary models fine-tuned for legal content and workflows.
By the numbers:
- $40 million — Thomson Reuters' investment in training the Thomson LLM
- 10,000 documents and 100 questions — capacity of CoCounsel Legal's Tabular Analysis feature
- 1 million users — number of professionals using CoCounsel across 107 countries
What's next: Details about wider implementation of Thomson across other Thomson Reuters products have not been announced yet.