Thomson Reuters Launches In-House AI Model, Challenges Frontier Model Economics
Event summary
- Thomson Reuters launched Thomson, its first proprietary large language model, developed in-house for $40 million.
- The model was trained on less than 10% of Thomson Reuters' proprietary content, with plans for further specialization.
- Thomson is integrated into CoCounsel Legal for high-volume document review, with plans to extend across legal and tax portfolios.
- A smaller version of Thomson is available as an open-weight model on Hugging Face for academic and non-commercial use.
The big picture
Thomson Reuters' launch of Thomson challenges the AI industry's focus on scale and compute-heavy models. By leveraging its proprietary content and domain expertise, the company aims to set a new standard for professional AI, emphasizing trust, accuracy, and sovereignty. This move positions Thomson Reuters as a key player in the evolving landscape of professional services AI, where domain-specific capabilities and cost efficiency are becoming increasingly important.
What we're watching
- Domain Specialization
- Whether Thomson Reuters can sustain its domain-specific gains over general-purpose models in professional tasks.
- AI Sovereignty
- The pace at which AI sovereignty becomes a critical differentiator for professional services firms.
- Cost Efficiency
- How Thomson Reuters' cost-efficient model development will impact the broader AI industry's economics.