Thomson Reuters Trained Its Own AI Model. Then It Kept Using Anthropic's Anyway.
The proprietary model is meant to cut reliance on outside AI vendors and lower heavy inference costs, with Thomson Reuters saying it spent $40 million.
- On Monday, Thomson Reuters launched Thomson-1, its first proprietary AI model, built by "realigning" an open-source Qwen model from Alibaba and internally codenamed Snowdon, Chief Technology Officer Joel Hron said.
- Thomson Reuters developed the model to leverage proprietary data rather than relying on third-party providers. Hron compared the strategy to buying a house: "you're not building any equity that compounds into something valuable for you long term" by renting.
- The company invested $40mn in talent and compute to train Thomson-1 on less than 10% of its proprietary content. Benchmarks show the model outperformed larger systems in three of seven categories, including legal reasoning tasks.
- Thomson-1 is deployed within Tabular Analysis in CoCounsel Legal, a feature capable of processing as many as 10,000 documents and answering up to 100 questions while verifying sources against Westlaw and Practical Law.
- This launch complements existing partnerships with Anthropic, as Thomson Reuters continues integrating third-party models where they fit best. Hron stated the objective is to "make Thomson the model that powers more and more of CoCounsel's capabilities over time.
13 Articles
13 Articles
Thomson Reuters built its own AI model on Alibaba's Qwen
Thomson Reuters launched its first proprietary large language model on Monday. The company calls it Thomson. It says it spent $40mn on talent and compute to train the model, and that it began from what its announcement describes as “a strong open-source foundation”. The announcement does not say which foundation. Its chief technology officer named […] This story continues at The Next Web
Here's how Thomson Reuters is loosening Claude's costly grip
Thomson Reuters CEO Steve Hasker.John Lamparski/Getty Images for Advertising Week New YorkThomson Reuters built Thomson-1 with a "realigned" version of Alibaba's Qwen3.5 model.Thomson-1 will save on costs, although it's not meant to replace Anthropic's Claude entirely.Companies are increasingly turning to open-source Chinese models as high AI prices bite.The Canadian data and legal giant Thomson Reuters wants to rely less on costly AI from Anthr…
Thomson Reuters trained its own AI model. Then it kept using Anthropic's anyway.
Thomson Reuters has developed its own AI model for legal, tax and compliance work, trained on the company’s proprietary professional content and designed to power features inside products such as CoCounsel. Early benchmarks show Thomson competing with models from OpenAI, Anthropic and Google across several professional and general-purpose evaluations. But Thomson wasn’t built from the ground up. The company started with an existing open-source f…
PR Newswire
Thomson Reuters Leverages its World-Class Data Assets to Launch Its Own Frontier Model
/PRNewswire/ -- Thomson Reuters (Nasdaq: TRI) (TSX: TRI), a global content and technology company, today announced the launch of Thomson, the company's first...
Thomson Reuters launches proprietary legal LLM "Thomson" - Legal IT Insider
Thomson Reuters today (24 August) announced the launch of Thomson, the company’s first proprietary large language model, leveraging an open-source foundation and investing $40 million to train the model. The […] The post Thomson Reuters launches proprietary legal LLM “Thomson” appeared first on Legal IT Insider.
Coverage Details
Bias Distribution
- 60% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
