

Thomson Reuters · 2× · last seen Aug 25, 2026
Thomson is Thomson Reuters' first proprietary, in-house Large Language Model, unveiled on August 24, 2026. It is based on an Open Source foundation model (reportedly Alibaba's Qwen, most recently Qwen 3.5), but has been intensively Fine-Tuned with proprietary content from Westlaw, Practical Law, Checkpoint, and Reuters, specifically targeting legal, tax, and compliance use cases. Its first commercial deployment is the Tabular Analysis feature in CoCounsel Legal for high-volume, structured document review, while CoCounsel remains a multi-model product. A smaller Open Source version is additionally made available on Hugging Face for research and non-commercial use.
Features
| Compliance/Certification | Designed to Thomson Reuters' "Fiduciary-Grade AI™" standard; customer data not used to train the model without explicit consent |
| Deployment Model | Cloud-based via CoCounsel Legal (multi-model platform); additional openly accessible small model version for external evaluation; sovereign AI options planned for the future |
| Use Case Scope | Legal, tax and professional services; first deployment: high-volume, structured document review (Tabular Analysis) |
| Integrations | CoCounsel Legal, Westlaw and Practical Law as training/tool environment; planned expansion across legal and tax product portfolio |
| License | Proprietary, fully owned by Thomson Reuters; small model variant under non-commercial academic license on Hugging Face |
| Platform | Integrated into CoCounsel Legal (Tabular Analysis); small version also available openly on Hugging Face |
| Price | No separate end-customer price disclosed; not sold standalone – accessed via CoCounsel Legal subscription; training cost about $40 million over 2 years (final training run about $450,000) |
| Release Date | August 24, 2026 (official launch) |