Cohere's Translate Model Beats DeepL and Google on WMT Benchmark

Cohere released North Small Translate and reported a WMT average score of 83.6, outperforming DeepL and Google Translate, while noting that translation remains a key AI challenge nine years after the Transformer paper.

2026-09-11 ~ 2026-09-11 · 4 related posts

• Cohere launches open translation model North Small Translate, scoring 83.6 on WMT — cohere · 2026-09-11
• North Small Translate hits 83.6 on WMT, outperforming DeepL and Google Translate — cohere · 2026-09-11
• Cohere says North Small Translate is proficient across 50+ languages — cohere · 2026-09-11
• Cohere open-sources North Small Translate model covering 50+ languages under CC BY-NC — cohere · 2026-09-11