North Small Translate

Skip to content AI for Empowerment: Your freedom. Your focus. See how AI gives you more time for what truly moves you. Explore now

Products

Solutions

Resources

Blog

Research

Company

Sign in

Request a demo

Platform North

Enterprise-ready AI for business

Compass

Intelligent search and discovery

Models Command

Generative language models

Transcribe New

Speech recognition model

North Small Translate NEW

Machine translation model

Parse New

Document parsing model

Embed

Semantic representation model

Rerank

Retrieval optimization model

Models Overview

Product Products Overview

Total Cost of AI Ownership

Pricing

Featured Command: High-performance generative AI models for real-world applications

Deploy Model Vault

Dedicated model inference platform

Private Deployments

On-prem or isolated VPCs

Security

Protect your data at every stage

See deployment options

By Industry Financial Services

Public Sector

Technology

Telecommunications

Energy and Utilities

Healthcare and Life Sciences

Manufacturing

Featured Model Vault provides fully-isolated, performant inference with Saas simplicity

Insights Customer Stories

For Developers Developers

Models Overview

Docs

Discord

LLM University

Connect Partners

Events

Webinars

Merch Store

Featured How CoreWeave used Cohere North to transform its customer support in 90 days

Blog

The latest news, launches, and insights

Read more

The state of sovereign AI adoption in 2026

Cohere and the University of Waterloo launch partnership to strengthen Canada’s AI talent pipeline

Introducing North Automations: Intelligent workflow orchestration

Research Cohere Labs

Cohere’s ML research lab

Explorations Future(s) of Work

How will AI change the way we work?

Aya Models

Multilingual AI at scale

All Papers

Initiatives Research Scholars

Finding the new generation of ML talent

Open Science Community

Championing global, open science

Catalyst Grants

Supporting impactful ML endeavors

Resources Blog

Hugging Face

Events

Featured The future of work debate has an evidence problem

About

Careers

Newsroom

Sep 10, 2026

4 minute read

Introducing North Small Translate: A leading sovereign open-weight machine translation model Outsized performance, right-sized footprint — translation built for speed and cost efficiency.

Today, we're releasing North Small Translate, a mixture-of-experts machine translation model with strong performance across 50+ languages. Across WMT26 benchmarks,¹ North Small Translate achieves an 83.6 score across all languages, outperforming proprietary models like DeepL and Google Translate, as well as open-weight alternatives such as Gemma 4 31B (off), GLM 5.2, and Mistral Large 3.

North Small Translate marks a significant milestone as Cohere's first translation model in the North model family. It builds on our multilingual and translation lineage — from the Tiny Aya model family to Command A Translate — and represents a clear next step in Cohere’s commitment to offering high-quality machine translation wherever it is needed.

Now available for research and non-commercial use under a CC BY-NC 4.0 license, North Small Translate advances Cohere’s mission to make sovereign AI a technological reality.

Visit Hugging Face to download the weights - available in several near-lossless quantizations - explore our HuggingFace space to demo the model, and read our implementation guides . Snapshot

Model North-Small-Translate-1.0

License Open-weights, non-commercial

Architecture MoE

Model size 218B total; 25B active

Context length 16k input, 16k output

Input modalities Text

Output modalities Text

Languages Supports 50+ languages. Full list

Optimized for Machine translation

Hardware (minimum) 1× B200 @ W4A4 2× H100s @ W4A4

Leading translation quality in the open ecosystem North Small Translate outperforms similarly sized open-weight models under 1T parameters and API-based translation models in various dimensions of machine translation on average. In WMT model evaluations, North Small Translate leads with an WMT26 All Languages benchmark score of 83.60, compared with 81.56 for Qwen 3.5 397B A17B, 76.50 for GLM 5.2 FP8, 81.37 for DeepL NextGen, 79.46 for Gemma 4 31B (on), and 68.20 for Google Translate. North Small Translate (Agentic) — which can find errors and fix errors in translation — scores even higher, at 84.36.² Image 1: Overall WMT26 scores across all languages across major models, benchmarked for WMT 2026 using GPT-5.6-Sol as a judge.

North Small Translate performs strongly across 32 high-resource languages and 18 additional languages. North Small Translate is the most consistent performer across the full spread of regions, without the sharp regional drop-offs seen in other models of its size. On average across all languages, it is the best-performing dedicated machine translation model in this evaluation — open or closed. Image 2: Overall WMT scores by regional average across DeepL, Google Translate, and Gemma 31B (on), and North Small Translate using GPT-5.6-Sol as a judge. At the regional level, North Small Translate punches above its weight, beating Gemma 4 31B (on) outright in Europe (82.2 vs. 73.9) while running essentially even with it in South Asia (86.2 vs. 86.7).

At the regional level, both North Small Translate and its Agentic counterpart beat Gemma 4 31B (on) outright across Europe — EU languages (82.74 Agentic / 82.17 standard vs. 72.73) and non-EU European languages (81.52 / 81.24 vs. 75.90) — while running essentially even with it in South Asia (87.13 / 86.16 vs. 88.04).

Both versions also outperform DeepL NextGen across every non-European region tested — MENA, South Asia, Southeast Asia, and East Asia — with the largest advantage in South Asia and MENA (roughly 8–10 points ahead of DeepL), a moderate edge in Southeast Asia (about 4–5 points), and the narrowest edge in East Asia (about 1-3 points, with the standard model closing in on DeepL's 85.41 score).

Increased throughput for faster workflows North Small Translate is built for high-throughput generation, prioritizing raw output speed even as concurrency scales.

In our testing, North Small Translate achieved up to 1.4x higher output throughput than Gemma 4 31B TP1 (1 x GPU) under identical concurrency levels and hardware configurations — 112 vs. 81 Output Tokens per Second (TOPS) at low concurrency and 39 vs. 30 TOPS at high concurrency. In practical terms, that's 30-38% more tokens...