The State of Translation Automation 2025

(Previously “State of Machine Translation”)

Be first to receive our annual report providing an independent multi-domain evaluation of Machine Translation engines and Large Language Models.

This year’s edition provides unique insights into the performance of leading Neural Machine Translation and Large Language Models across full-text translations, carefully evaluated against essential business criteria. We’re excited to share practical findings that can directly enhance your multilingual strategy and user experience.

MT Engines and LLMs

Language pairs

Content Domains

MT Engines and LLMs

Language pairs

Content Domains

Trusted by global enterprises

What‘s changed in the MT market since 2023

LLMs reshaped the MT landscape and now account for 55% of top-performing models
Quality varies by language and domain, with LLMs excelling in Colloquial, Education, and Entertainment
Traditional Machine Translation (MT) still outperforms in specific language pairs and domains
LLMs are less expensive but slower than MT engines
Open-source LLMs lag behind commercial offerings
Customization through translation memories, glossaries, and prompt engineering is key to eliminating errors

52 Evaluated MT Engines and LLMs

MT Engines

LLMs

MT Engines

LLMs

18 providers show the best results and 9 MT Engines excel among real-time providers

18 providers show the best results

Our datasets cover:

22 language pairs (English-to-Spanish, French, Italian, Portuguese, German, Dutch, Ukrainian, Korean, Japanese, Chinese, and Arabic)

9 content domains (General, Colloquial, IT, Entertainment, Hospitality, Education, Healthcare, Legal, and Financial)

9 MT Engines excel among real-time providers

“We launched our inaugural MT report in 2017, at a time when there were only eleven stock models available. Fast forward six years, and the MT landscape has exploded, with six-fold growth in MT platforms and 15x more language pairs. In addition, we see new automated customization possibilities such as domain adaptation and custom glossaries, and the emergence of Large Language Models, which can handle translation among other tasks.”

“We’re pleased to extend our expertise to drive exceptional MT evaluation results in the latest Intento report. The caliber of our lingosets (golden datasets) is backed by a talented force of translation and data scientists that have enabled e2f to partner with some of the top players in the localization and AI landscape. As a result, enterprises are attaining a new standard of excellence for conversational AI.”

The State of Translation Automation 2025

What‘s changed in the MT market since 2023

52 Evaluated MT Engines and LLMs

Alibaba Cloud

Alibaba

Amazon

Baidu

DeepL

Elia

HiThink

Globalese

Google Cloud

Kawamura by NICT

Lingua Custodia

Microsoft

Mirai

ModernMT

Naver

NiuTrans

Oracle

Pangeanic

PROMT

SYSTRAN

Tarjama

TartuNLP

Tencent Cloud

Tilde

TREBE

Ubiqus

Yandex

Youdao

Aya-101

Claude 3 Haiku

Claude 3 Opus

Claude 3 Sonnet

Command-R

Command-R+

Gemini Pro

Gemini Pro 1.5

GPT-3.5 Turbo

GPT-4

GPT-4о

GPT-4о

GPT-4 Turbo

AI21

LLaMA-2

LLaMA-3

Mistral Large

Mixtral 8x7B

GPaLM2 Chat Bison

PaLM2 Text Bison

PaLM2 Unicorn

RakutenAI 7B Instruct

Titan Text Express

TowerInstruct 13B v0.1

TowerInstruct 7B v0.2