- RWS TrainAI study found multilingual performance gap between English and underrepresented languages has narrowed across leading large language models.
- Google Gemini Pro scored above 4.5 out of 5 in Kinyarwanda, a language where earlier model generations struggled to produce coherent text.
- Research flagged “benchmark drift,” with capabilities shifting unpredictably between model releases, including latest GPT version trailing smaller models on several content-generation tasks.
- Tokenizer efficiency varied sharply by model and language, with cost differences reaching 3.5 times in certain languages.
- Report urged enterprises to run continuous, independent evaluations on each new model release rather than rely on public leaderboards.
Disclaimer: This news brief was created by Public Technologies (PUBT) using generative artificial intelligence. While PUBT strives to provide accurate and timely information, this AI-generated content is for informational purposes only and should not be interpreted as financial, investment, or legal advice. RWS Holdings plc published the original content used to generate this news brief via Business Wire (Ref. ID: 202604130401BIZWIRE_USPR_____20260413_BW820932) on April 13, 2026, and is solely responsible for the information contained therein.
Comments