Building Multilingual Bridges

In-Language Reasoning: Data Mixing Bridges | Cohere Skip to content Introducing North 2: AI without compromises Learn more about the updates

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

North 2: Enterprise AI without compromises

Introducing Embed 5—A new family of frontier embedding models

Cohere and Aleph Alpha sign agreement to become the first transatlantic sovereign AI solution

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

Oct 06, 2026

9 minute read

Multilingual Bridges: How Data Mixing Unlocks In-Language Reasoning When a reasoning model gets an answer wrong, you’d read its reasoning to find out why. But what if that reasoning is in a language you don't speak?

Key takeaways Truly multilingual AI should reason in the user's language, not just answer in it. Today's reasoning models mostly think in English, which can lose nuance from the prompt in internal translation, leave language-specific cultural knowledge unused, and exclude non-English speakers from inspecting or engaging with the reasoning traces. Moving from multilingual answers to in-language reasoning (referred to as “L2” reasoning) is the next step toward reasoning systems that are genuinely accessible across different languages. We find that in-language reasoning can be achieved through data mixing built on three pillars. First, ensuring a broader language coverage in reasoning training data improves generalization to languages without reasoning supervision. Second, including a small share of multilingual non-reasoning data boosts both accuracy and in-language reasoning. And third, using English reasoning data provides the task-solving backbone. Our resulting model, Tiny Aya L2-Thinker (3.35B), reasons in the prompt language over 93% of the time across 60 languages. It does so with minimal accuracy loss, holds up on low-resource languages, and uses fewer reasoning tokens than comparable models. A model’s ability to reason and the language it reasons in are not always tied to one another, so L2 reasoning doesn't have to be built language by language. This puts in-language reasoning within reach even where multilingual reasoning data is scarce. We release the model weights and multilingual reasoning data to support further work on accessible, in-language reasoning.

The language of reasoning matters Some of the biggest recent gains in language models have come from a simple idea: let the model ‘think’ before it answers. The result of this idea is the development of reasoning models that work through a problem step by step, producing a written ‘reasoning trace’. This has led to great improvements in language models’ abilities to complete tasks like math and coding. These models, however, mainly focus on giving the right answer, but we ask an often-overlooked question: can users actually follow the reasoning traces? For many users, the answer today is no: when a question is asked in Spanish, Arabic, or Swahili, these models tend to reason in English. The final answer may come back in the user's language, but the reasoning steps behind that answer stay out of reach for anyone who doesn't speak English. This is not an unsolvable problem. By enabling multilingual reasoning models to reason in the user’s language, which we call ‘ in-language ’ or ‘ L2 ’ reasoning, users can read the reasoning trace, verify each step, find where things went wrong when the answer is incorrect, and step in when needed. L2 reasoning may also help the model itself. Much knowledge, such as cultural context and linguistic nuance, is expressed most naturally in its source language. L2 reasoning could let models draw on this culturally embedded knowledge more effectively, especially in linguistic and open-ended generation tasks. The example below shows what this looks like in practice: Asking in Hindi which event signals good luck, our Tiny Aya L2-Thinker model reasons in Hindi about what Indian superstitions hold and reaches the correct answer, while thinking in English asserts a Western-biased answer without weighing the options. In this blog post, we describe our recent research that shows how better data mixing leads to effective L2 reasoning without hurting accuracy much. Building on the Tiny Aya base model with 32K context length , we apply our data mixing strategy to create Tiny Aya L2-Thinker , a massively multilingual model achieving over 93% in-language reasoning rate across 60 languages. What’s missing from prior work Getting models to reason in the user's language is not a new idea, but earlier attempts have fallen short in three ways. First, they show that L2 reasoning often comes at a cost: models that are pushed to reason in another language tend to get more answers wrong , and approaches that avoid this drop rely on complex, expensive training ....