Cohere's North Small Translate machine translation model sets a new standard, blending extensive language capabilities with economic efficiency. Leveraging an advanced mixture-of-experts (MoE) architecture with 218 billion total parameters, of which 25 billion are active, the model illustrates an impressive balance between capability and resource utilization. Specifically designed for machine translation, North Small Translate supports over 50 languages, offering a robust solution for multinational enterprises demanding precision in numerous languages.

On the performance front, North Small Translate scores an impressive 83.60 across all languages on the WMT26 benchmark, which effectively situates it ahead of many proprietary models, including industry giants like DeepL and Google Translate. Its competitive edge becomes even clearer in regional applications, such as European Union languages, where it achieves a score of 82.74, markedly surpassing Gemma 4 31B’s 72.73. Furthermore, its long-context evaluation score of 48.9 dramatically outpaces Google's 21.3, highlighting its superiority in processing longer, more complex translation tasks. This exceptional performance is coupled with efficient throughput, outperforming peers such as Gemma 4 31B in both low and high concurrency scenarios, processing 112 and 39 output tokens per second, respectively.

Cost efficiency is another domain where North Small Translate excels. As reported by Cohere, the cost per task for commercial licenses is remarkably low at $0.000676, making it an attractive option for organizations with substantial translation demands. This economic advantage is achieved without compromising on translation quality, aligning with Cohere's commitment to offering high-performance models that remain financially viable. The average task translates around 661 tokens, which means that even large-volume projects remain manageable within budget constraints.

In summary, North Small Translate embodies the confluence of innovative design and practical application, offering a multilingual machine translation solution that is both powerful and cost-efficient. Its availability on platforms like Hugging Face for non-commercial use positions it as a versatile tool for researchers and developers dedicated to advancing translation technologies further. As businesses continue to seek out translation solutions that are both precise and scalable, North Small Translate offers a compelling proposition by demonstrating that true performance does not have to come at a prohibitive cost.