The AIM-listed language technology company says its new model, the largest dedicated translation model in production, ranked first in 31 of 32 languages in internal benchmarking against DeepL and Gemini. The model is natively integrated into the Trados portfolio.
The problem RWS is describing is a real one in enterprise translation: the tools that move fastest, general-purpose AI models that happen to translate, tend to produce output that is fluent but contextually imprecise, inconsistent between passes, and built for consumer use cases rather than regulated industries.
A legal contract, a pharmaceutical label, or a government policy document has a very different tolerance for ambiguity than a marketing email. RWS, the UK-listed language technology and AI solutions company, has launched Language Weaver Pro, a new enterprise translation model built to close that gap.
Language Weaver Pro is a 100-billion-plus parameter model built in partnership with Cohere, the Canadian enterprise AI company founded in 2019 by Aidan Gomez.
RWS describes it as the largest dedicated translation model currently in production, and says internal benchmarking, using both human-led and automated evaluation across sentence and paragraph-level datasets covering factual and marketing content, placed Language Weaver Pro first in 31 of 32 languages against DeepL and Google Gemini.
The model is available immediately and is natively integrated into the Trados portfolio, RWS’s translation management platform, so enterprise teams can access it within the localization workflows they already use.
The Cohere partnership, which RWS’s Q2 2025 earnings call described as exclusive, brings together two complementary capabilities: Cohere’s security-first frontier AI model infrastructure, and RWS’s decades of linguistic fine-tuning data, domain expertise, and enterprise integration experience.
The collaboration is branded as RWS’s Language Intelligence capability, a broader positioning that encompasses the new model’s underlying infrastructure alongside the governance and compliance tooling that enterprise and government customers require.
Gomez, Cohere’s CEO and co-founder, framed it as an infrastructure play: “High-quality translation is essential infrastructure for global businesses operating across borders.”
The competitive positioning against DeepL and Gemini is the most directly provocative element of today’s announcement. DeepL has built a substantial enterprise translation business on the argument that dedicated translation models outperform general-purpose AI.
RWS is essentially making the same argument but against both DeepL and the general-purpose incumbents simultaneously, claiming a larger model trained on more specialised data.
RWS was founded in 1958 and has been listed on AIM (RWS.L) since 1997. It employs around 250,000 data specialists, cultural experts, and language professionals across its network and holds 45-plus AI-related patents. The company serves more than 80 of the world’s top 100 brands, including most of the top pharmaceutical companies and patent filers.
Ben Faes, who became CEO in January 2025 after running Google Cloud for Southern Europe and Emerging Markets, is driving the company’s pivot from traditional translation services to AI-first solutions. Language Weaver Pro is the most significant product expression of that strategy to date.
