The AIM-listed language technology company has unveiled its new model, the largest dedicated translation model in production, which ranked first in 31 of 32 languages in internal benchmarking against DeepL and Gemini. It is natively integrated into the Trados portfolio.
RWS addresses a common issue in enterprise translation: general-purpose AI models that translate rapidly often result in output that is fluent but contextually inaccurate, inconsistent, and more suited for consumer use cases than regulated industries. Legal contracts, pharmaceutical labels, or government policy documents require more precision. RWS, a UK-listed language technology and AI solutions company, has introduced Language Weaver Pro, an enterprise translation model designed to address this challenge.
Language Weaver Pro is built in collaboration with Cohere, a Canadian enterprise AI company founded in 2019 by Aidan Gomez, and features a model with over 100 billion parameters. According to RWS, it is the largest dedicated translation model currently in production. Internal benchmarking, through both human-led and automated evaluations across sentence and paragraph datasets covering factual and marketing content, showed Language Weaver Pro performing best in 31 of 32 languages when compared to DeepL and Google’s Gemini.
The model is immediately available and is integrated into RWS’s Trados portfolio, allowing enterprise teams to access it within their existing localization workflows. The collaboration with Cohere, described as exclusive in RWS’s Q2 2025 earnings call, combines Cohere’s security-first AI model infrastructure with RWS’s data, domain expertise, and integration experience, collectively branded under RWS’s Language Intelligence capability. This broader positioning includes the model’s infrastructure and the necessary governance and compliance tools for enterprise and government customers.
Cohere’s CEO and co-founder, Aidan Gomez, described high-quality translation as essential infrastructure for global businesses. The announcement’s competitive positioning against DeepL and Gemini is notable, with RWS claiming a larger model trained on more specialized data against both competitors.
Founded in 1958 and listed on AIM since 1997, RWS employs around 250,000 data specialists and holds over 45 AI-related patents. It serves more than 80 of the world’s top 100 brands, including leading pharmaceutical companies and patent filers. Ben Faes, who became CEO in January 2025 after working with Google Cloud, is leading the company’s shift to AI-first solutions, with Language Weaver Pro being the most significant product result of this strategy.