Trusted AI built on 175 years of Thomson Reuters knowledge. Meet The CoCo

Meet Thomson

Thomson Reuters doesn’t just deploy AI — we build it

Introducing Thomson, our proprietary LLM, purpose-built for legal, tax, and regulatory work where accuracy, consistency, and auditability aren't optional

Purpose-built for professionals, tested by academics

Until now, the most capable AI models have come from frontier labs that focus only on generalized capabilities for any user. Thomson doesn’t — it starts from a strong, open-source foundation, then trains on decades of authoritative Thomson Reuters proprietary legal, tax, and regulatory content that no general-purpose model can touch. The result is a model that competes with the best frontier models in the world and is purpose-built for professional work.

“Thomson is competitive with the world's leading frontier models despite being a fraction of their size and cost to train and operate.”

Joel Hron
Chief Technology Officer, Thomson Reuters

A foundational model that ranks among the world’s best

Thomson is evaluated against more than 100 legal-domain benchmarks and thousands of legal-specific use cases, with independent, academic benchmarking underway with partner institutions.

Frontier-model competitive

Thomson performed competitively with the strongest frontier models on the market, including Claude Opus 4.8, and ahead of GPT-5.5, Claude Sonnet 5, and Gemini 3.1 Pro.

#1 on the hardest legal benchmark

With a top score of 0.352 on PrBench Legal Hard, Thomson is ahead of every frontier model tested.

Ready to train on even more content

Less than 10% of our decades of proprietary content has been used in training so far — the rest is fuel for what comes next.

Model benchmark comparison

As we leverage more of our proprietary legal, tax, and regulatory content within the model, we reevaluate its performance against the largest general-purpose models in the world. Our most recent benchmarks show Thomson performs competitively with frontier models on some of the most important professional, legal, and general reasoning tasks.

*Head-to-head comparison across Legal Domain and General Domain benchmarks; the best score in each row is highlighted in green

How we built it

Technical report

Our technical report, with our experimental findings and details of the methods, data, and evaluations from training the Thomson model, is available here.

“Our evaluation found Thomson’s citation quality generally competitive with leading frontier models, even when tested on Canadian employment-law questions without a Canada-specific setting.”

Professor Samuel Dahan, Director, Queen's Conflict Analytics Lab and Cornell Legal AI Lab

“I tested Thomson against ChatGPT and Claude using some of the more challenging questions students have asked in my Corporate Tax class. All three models answered the questions correctly, but I preferred Thomson's responses overall. I especially appreciated the links to treatises, which made its responses more transparent and useful for legal work.”

Jonathan H. Choi, Washington University School of Law

Thomson in action: Proof in production

The first way customers will experience Thomson

The first place we're using Thomson is inside Tabular Analysis in CoCounsel Legal. Tabular Analysis reviews up to 10,000 documents against up to 100 questions at once, with every answer traceable back to its source. It's exactly the kind of high-volume, judgment-heavy review where a purpose-built model's edge is measurable, not just claimed. Plus, this is just the start — we have plans to expand Thomson across more of our portfolio soon.

Your data secured — by architecture, not just by policy

We do not train Thomson on customer data. Your data is yours and stays that way, with the assurance that only Thomson Reuters can determine how Thomson is trained, where it runs, and how it behaves. For more information, please access the "CoCounsel Legal Next-Gen" security white paper found in our Trust Center.

Frequently Asked Questions

What is Thomson?

Thomson is our proprietary large-language model, purpose-built for professional work. It's built on open-source foundations, then trained on decades of our proprietary case law, contracts, tax content, and Reuters news, validated by in-house domain experts.

How does Thomson fit into the Thomson Reuters portfolio?

Thomson is one of the model layers running underneath our products — starting with Tabular Analysis in CoCounsel Legal and soon expanding across the portfolio. It's one part of a multimodel strategy, not a replacement for the products themselves.

How is Thomson different than CoCounsel?

CoCounsel is the product. Thomson is the model that powers specific capabilities within our products, starting with Tabular Analysis. One is the experience. The other is the engine that powers it.

Can customers buy or access Thomson directly today?

No. We do not currently price, sell, or offer Thomson as a standalone model.

Where will customers experience Thomson first?

Customers can soon experience Thomson through CoCounsel Legal's Tabular Analysis capability.

Does Thomson train on customer data?

No. We do not train Thomson on customer data.

Does Thomson replace other models Thomson Reuters uses?

We use a multimodel strategy and apply the right model to the task within each product.

Want to keep learning about what Thomson can do for your work?

Fill out the form to receive future updates as they become available.

 

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.