Thomson Reuters builds $40M legal LLM rivaling Claude Opus on Qwen
josh_wills · x · 2026-09-01
Thomson Reuters has launched "Thomson," a custom LLM built on Qwen3.5-397B and fine-tuned on 175 years of proprietary legal, tax, and news data.
Key Details:
- Cost & Strategy: While the total investment was $40M, the model training cost only $450k, with the bulk spent on the "model factory" infrastructure. This allows them to swap in new open-source base models to stay at the frontier.
- Performance: On internal benchmarks, it reportedly matches Claude Opus 4.8 and outperforms GPT-5.5, Claude Sonnet 5, and Gemini 3.1 Pro.
- Team: Built by the team from Safe Sign, a pre-revenue legal AI startup acquired two years ago.
- Data: Less than 10% of their data collection has been utilized so far.
This case highlights a trend where frontier-level custom models become the standard for enterprises.
Related event: Thomson Reuters Unveils $40M Legal LLM(2 posts)→
More from Companies & People
- Tinyfool sparks debate over whether Tesla Roadster 2 counts as 'useless' innovation — tinyfool · 2026-10-16
- Tinyfool: does Xiaomi's foldable released days before iPhone count as innovation? — tinyfool · 2026-10-16
- US universities go on an AI philosopher hiring spree as philosophy departments add tenure-track AI roles — AnnaCiaunica · 2026-10-16
- NVIDIA and Mercedes-Benz Expand Partnership on Next-Gen Automated Driving — nvidia · 2026-10-16
- Ex-Amazon Engineer: AI Has Killed Bezos's Six-Page Memo Culture — bibryam · 2026-10-16
- DHH's Omarchy Shenzhen Meetup Draws 400+ With Demos on Xiaomi, ARM and iPhone Ports — LoganGrasby · 2026-10-16