EfficientXLang: Towards Improving Token Efficiency Through Cross-Lingual Reasoning

TL;DR

This paper demonstrates that reasoning in non-English languages can be more token-efficient and equally accurate compared to English in multilingual language reasoning models, emphasizing the value of multilingual reasoning.

Contribution

It provides empirical evidence that multilingual reasoning reduces token usage without sacrificing accuracy, highlighting the importance of multilingual foundations in language models.

Findings

Non-English reasoning reduces token usage.

Accuracy is preserved across languages.

Gains depend on multilingual model strength.

Abstract

Despite recent advances in Language Reasoning Models (LRMs), most research focuses solely on English, even though many models are pretrained on multilingual data. In this work, we investigate: Is English the most token-efficient language for reasoning? We evaluate three open-source RLMs: DeepSeek R1, Qwen 2.5 and Qwen 3, across four math datasets and seven typologically diverse languages. We find that reasoning in non-English languages not only reduces token usage, but also preserves accuracy. These gains persist even after translating the reasoning traces into English, suggesting genuine shifts in reasoning behavior rather than surface-level linguistic effects. The extent of improvement, however, depends on the models multilingual strength. Our findings motivate a broader view of reasoning in language models, highlighting the potential of multilingual reasoning and the importance of strong multilingual foundations. The code for our work can be found: https://github.com/microsoft/EfficientXLang.