Cohere releases open-weight North Small Translate covering 50 languages

single source · 1 articles · release · confidence: high · first seen 2026-09-11 06:57 UTC

Cohere has released North Small Translate, an open-weight mixture-of-experts model for machine translation across 50 languages. It uses 25B of its 218B parameters per token and scores 83.6 on Cohere's WMT26 evaluation. Weights are free for non-commercial use; commercial access is through Cohere Model Vault or RWS Language Weaver.

What this means for you

If you need machine translation across many languages, evaluate this against your own pairs. It is free for non-commercial use, but commercial deployment requires a licence through Cohere Model Vault or RWS Language Weaver.

Key facts

  • ·North Small Translate is an open-weight mixture-of-experts model with 218B total parameters and 25B active per token. source
  • ·The model covers 50 languages. source
  • ·It scores 83.6 on Cohere's WMT26 evaluation. source
  • ·Weights are free for non-commercial use. source
  • ·Commercial access is through Cohere Model Vault or RWS Language Weaver. source

What the sources say

  • MarkTechPostReports the model's release, architecture, benchmark score, and licensing terms.

Sources

The original reporting. Follow these — they did the work.

← the wire