Thomson Reuters Corporation is a content and technology company listed on Nasdaq and the Toronto Stock Exchange under the ticker TRI. It sells information products and workflow software to legal, tax and accounting, risk and fraud, trade, and government customers, and owns the Reuters news agency. The company describes itself as "an AI and technology company empowering professionals with trusted content and workflow automation," and dates its content franchise to "over 150 years" (Source: thomsonreuters.com). Its AI products are marketed under the CoCounsel brand and a company standard it calls Fiduciary-Grade AI. In July 2026 it disclosed benchmark results for Thomson, a legal-domain large language model it trained itself, and in August 2026 deployed that model inside CoCounsel Legal.
Products
| Product | Description |
|---|---|
| Westlaw / Westlaw Edge / Westlaw Advantage | Legal research platform; Westlaw Advantage adds agentic AI research over verified Westlaw content |
| Practical Law | Attorney-editor-written guides, templates and checklists |
| CoCounsel Legal | AI assistant for legal research, drafting, analysis and verification |
| CoCounsel Tax / CoCounsel Audit & Accounting | AI assistants for tax research and audit workflow |
| Checkpoint | Tax and accounting research service |
| CLEAR | Public-records investigation and risk-screening platform |
| ONESOURCE | Corporate tax, indirect tax and global-trade compliance software |
| Reuters | News agency and media products |
(Source: thomsonreuters.com)
Thomson (proprietary model)
Thomson is a large language model developed in-house and directed at professional legal and tax work. Chief Technology Officer Joel Hron and Head of AI Research Jonathan Schwarz published first benchmark results on July 31, 2026, describing the model as starting "from a strong open-source foundation" and then trained with mid-training and post-training techniques on Westlaw, Practical Law, Checkpoint and Reuters content, with output evaluation by "hundreds of subject matter experts." They state that customer data is never used to train the model, and that fewer than 10% of the company's content holdings had been used in training to that point (Thomson Reuters Built Its Own AI Model That Now Ranks Among the World's Best).
The company traces the model to its 2024 acquisition of Safe Sign Technologies, an AI research company, and frames the build as a bet that "the future of professional AI would require more than general-purpose intelligence" (Thomson Reuters Built Its Own AI Model That Now Ranks Among the World's Best).
Reported benchmarks
The July 2026 post published a head-to-head table for the variant it labels Thomson-1-Large against Gemini 3.1 Pro, Claude Opus 4.8 and GPT-5.5. The company notes that Thomson-1-Large used test-time scaling, that Gemini 3.1 Pro and Opus 4.8 used reasoning mode, and that GPT-5.5 used non-reasoning mode — so the comparison is not like-for-like across inference settings. Scores are self-reported and had not received independent validation at publication. The post's own summary claim — that Thomson performed ahead of GPT-5.5, Claude Sonnet 5 and Gemini 3.1 Pro — is broader than the table supports: Claude Sonnet 5 does not appear in it, and Thomson-1-Large leads each of GPT-5.5 and Gemini 3.1 Pro on four of the seven rows.
| Benchmark | Thomson-1-Large | Gemini 3.1 Pro | Opus 4.8 | GPT-5.5 |
|---|---|---|---|---|
| Stanford LegalBench | 0.823 | 0.843 | 0.818 | 0.832 |
| PrBench Legal Hard | 0.352 | 0.293 | 0.315 | 0.333 |
| Harvey Legal Agent Benchmark | 0.857 | 0.555 | 0.869 | 0.781 |
| Instruction Following (IFEval + FollowBench) | 0.914 | 0.848 | 0.861 | 0.885 |
| Reasoning (GPQA Diamond, HLE, MMLU-Pro) | 0.684 | 0.748 | 0.737 | 0.589 |
| Coding (SWE-Bench Pro, Terminal-Bench 2.1) | 0.399 | 0.500 | 0.598 | 0.414 |
| Long Context (Infinity Bench + internal) | 0.753 | 0.750 | 0.741 | 0.703 |
(Thomson Reuters Built Its Own AI Model That Now Ranks Among the World's Best)
A second evaluation covered 53 legal research queries written by internal subject-matter experts, with models connected through an in-house agentic harness to Westlaw and Practical Law for company content and to the Brave search engine for web search. Completeness and factuality were scored by language models acting as judges against expert-written rubrics (Thomson Reuters Built Its Own AI Model That Now Ranks Among the World's Best).
Deployment and cost
Thomson entered service on August 24, 2026 in the Tabular Analysis document-review capability of CoCounsel Legal, becoming the default model there while administrators retain the ability to select others. The company said it spent about $40 million over two years on people and computing for the project, and that efficiencies brought the final training run to about $450,000. Hron, described in that account as global head of artificial intelligence and TR Labs, said Thomson "needs to set the frontier of intelligence for legal," which he distinguished from the job of the frontier labs. Senior research scientist Andrew Bean said internal tests showed the model broadly competitive with leading models when all had web access only, and roughly equal or slightly better when connected to Thomson Reuters content; a technical report was said to be expected later. A smaller open-weight version is planned for release on Hugging Face under a noncommercial academic license (Source: siliconangle.com).
CoCounsel Legal
Thomson Reuters announced general availability of the next generation of CoCounsel Legal on August 20, 2026, describing it as an agentic AI experience spanning research, drafting, legal intelligence, verification and matter-centric workflows. Raghu Ramanathan, President, Legal Professionals, said the challenge for legal organizations is "turning insight into action." The company states the product is built on Anthropic's Claude Agent SDK and that one million professionals across 107 countries and territories use CoCounsel (Source: thomsonreuters.com).
Capabilities announced with that release include Tabular Analysis, powered by Thomson; Deep Research Verify, which checks whether cited Westlaw and Practical Law authority supports specific legal assertions; and Westlaw Brief Builder, which moves from research and issue analysis to first-draft briefs. Emily Colbert, SVP, CoCounsel Litigation, said the tool "was built to support that thinking, not replace it" (Source: thomsonreuters.com).
Partnerships
The company announced an expanded CoCounsel Legal Model Context Protocol server for Claude in August 2026, allowing legal professionals to reach CoCounsel Legal from within Claude and receive cited work product grounded in Westlaw, Practical Law and customer content. It also announced a forthcoming MCP connection extending CoCounsel Legal into Amazon Quick and other AWS offerings, and an integration with the e-discovery vendor Reveal beginning in October 2026 (Source: thomsonreuters.com).
Market position
CoCounsel is described in coverage of the legal-AI market as the incumbent offering against which venture-funded entrants position themselves, including Harvey and Legora, alongside firm-built systems such as Springbok AI at Cleary Gottlieb. Adoption patterns across the sector are tracked at Legal Services — AI Deployment.
Relationships
- deploys-in: Legal Services — AI Deployment — CoCounsel and Westlaw are sold into law firms, corporate legal departments and government legal offices.
- related: Anthropic — CoCounsel Legal is built on Anthropic's Claude Agent SDK, and the two companies shipped an MCP integration in August 2026.
- related: Amazon — a forthcoming MCP connection is to extend CoCounsel Legal into Amazon Quick and other AWS offerings.
- related: Harvey, Legora — competing legal-AI vendors.
- related: Thomson (Thomson Reuters legal-domain model) — the company's in-house legal-domain model.
- instance-of: Professional Services AI Adoption — a professional-information incumbent building a domain-specific model rather than relying on general-purpose ones.