# Thomson Reuters enters the model race with Thomson-1.0

> The information giant's first frontier model is a mixture-of-experts system tuned on its proprietary legal, tax, and news data.

Published by The Open Weights on Aug 18, 2026. Canonical: https://theopenweights.com/news/thomson-1-0-small-i9ut

## Key facts

- Company: Thomsonreuters
- Model: Thomson-1.0-Small
- Version: 1.0 Small
- Category: Text / LLM
- Modalities: Text / LLM, Vision-Language
- License: Other (Open weights, commercial use allowed)
- Context window: 262K tokens
- Architecture: Qwen3_5MoeForConditionalGeneration
- Precision: BF16
- Weights size: 70.2 GB
- Significance: notable
- Published: 2026-08-18
- Last verified: 2026-09-04
- Hugging Face: https://huggingface.co/thomsonreuters/Thomson-1.0-Small
- Canonical URL: https://theopenweights.com/news/thomson-1-0-small-i9ut

Thomson Reuters has released its first frontier model, [Thomson-1.0-Small](https://huggingface.co/thomsonreuters/Thomson-1.0-Small), a mixture-of-experts (MoE) language model the company says is tuned on its proprietary data assets. It marks the moment a legacy information company decided that owning the model, not just the content, is worth the investment.

The pitch is straightforward: Thomson Reuters sits on decades of curated legal, tax, and news material through franchises like Westlaw and Reuters. According to the company's [announcement](https://www.thomsonreuters.com/en/press-releases/2026/august/thomson-reuters-leverages-its-world-class-data-assets-to-launch-its-own-frontier-model), Thomson-1.0-Small is designed to put that proprietary corpus to work directly inside a model rather than bolting it on through retrieval alone.

## What's in the release

- A mixture-of-experts architecture, which routes tokens to specialized sub-networks to keep inference efficient relative to total parameter count.
- Listed capabilities spanning text generation and vision-language understanding, with text as the primary modality.
- A custom license rather than a standard permissive one, so teams will need to read the terms before building on it.

The company has not published parameter counts, context length, or benchmark figures alongside the initial release, so its standing against general-purpose open models remains an open question. The name "Small" also implies a larger sibling may follow.

Why it matters: data owners have long licensed their archives to AI labs, often uneasily. Thomson Reuters shipping its own model is a signal that some of those companies would rather compete than supply — and that domain-specific data may be the clearest moat left in a crowded model market.

## Get the model

- [Hugging Face](https://huggingface.co/thomsonreuters/Thomson-1.0-Small)

## Sources

- [thomsonreuters/Thomson-1.0-Small](https://huggingface.co/thomsonreuters/Thomson-1.0-Small) — Hugging Face, Aug 18, 2026

---
Source: The Open Weights (https://theopenweights.com/). Aggregated and written by Claude, curated by humans. Cite as: "Thomson Reuters enters the model race with Thomson-1.0", The Open Weights, Aug 18, 2026, https://theopenweights.com/news/thomson-1-0-small-i9ut