Thomson Reuters Launches Proprietary Large Language Model 'Thomson'

Thomson Reuters has announced the launch of 'Thomson,' its first proprietary large language model (LLM), developed in-house. The company invested $40 million to train the model, leveraging a strong open-source foundation and decades of proprietary content from sources like Westlaw, Practical Law, Checkpoint, and Reuters. This approach aims to create a highly capable and efficient AI at a fraction of the typical cost associated with frontier models, providing Thomson Reuters full control over the model and reducing heavy inference costs.

Context

Thomson Reuters, a major player in legal and financial information services, has invested $40 million to create 'Thomson,' its proprietary large language model. The model is built on a strong open-source foundation and utilizes decades of proprietary content, allowing for advanced capabilities tailored to the company's needs. This development reflects a broader trend among companies seeking to harness AI technology for specialized applications.

Why it matters

The launch of 'Thomson' represents a significant step for Thomson Reuters in the competitive AI landscape. By developing its own large language model, the company aims to enhance its offerings in legal and financial services. This move could set new standards for efficiency and cost-effectiveness in AI applications within these sectors.

Implications

The introduction of 'Thomson' could lead to increased competition among AI providers in the legal and financial sectors. It may also influence how organizations approach AI investments, particularly regarding cost and control. Companies relying on Thomson Reuters' services may benefit from improved efficiency and tailored solutions, while competitors may need to innovate to keep pace.

What to watch

In the near term, stakeholders should monitor how 'Thomson' performs in real-world applications, particularly in legal and financial contexts. The company's ability to integrate this model into its existing platforms will be crucial. Additionally, reactions from competitors and the market will provide insights into the model's impact on the industry.