Thomson Reuters Spent $40M Training Its Own LLM on a Qwen Base
FinanceYF5 · x · 2026-09-04
Thomson Reuters reportedly spent $40 million training its own LLM, Thomson, built on a Qwen base and continuously trained on 175 years of legal, tax, accounting, and news data. Official benchmarks place it near Claude Opus, ahead of GPT and Gemini rivals — and the company says it has used less than 10% of its data so far. A striking case of vertical-domain LLMs built on open-weight foundations plus proprietary data.
More from Companies & People
- 43 Projects From Singapore's ChatGPT Sites Hackathon Open for Voting — gabrielchua · 2026-09-04
- Microsoft Research launches internship series starting with agentic systems research — ZacharyHuang12 · 2026-09-04
- Nebius hits $30B market cap in 2 years, nearly 3x Yandex peak, says analyst — robleclerc · 2026-09-04
- Musk calls for a 'construction army' to staff xAI's Memphis supercomputer buildout — elonmusk · 2026-09-04
- Dev slams OpenAI's tiered rollouts as betrayal of its founding 'open' mission — ctjlewis · 2026-09-04
- 9 resume rules to pass hundreds of dev screens, from ditching AI slop to 3 deep projects — _jaydeepkarale · 2026-09-04