The State of Translation Automation 2026Performance

Two-step solutions take ~twice as long as the same solutions in one step, and three-step solutions ~three times as long.

← back

In German, Bosnian, Georgian and Albanian, a two-step solution on a faster LLM is quicker than the fastest one-step solution that meets every Standard requirement. In German, the two-step solution on OpenAI GPT-5.6 Luna takes 30 seconds per 1,000 words, against 79 seconds per 1,000 words for the fastest one-step solution.

The fastest solution that meets every Standard requirement, by language

LanguageStepsModels per 1,000 wordsFastest one-step solution, s per 1,000 words
Afrikaans3Anthropic Claude Fable 5203.1
Albanian2OpenAI GPT-5.6 Terra68.885.2
Amharic3Google Gemini 3.1 Pro Preview316.2
Arabic (MSA)1Anthropic Claude Fable 557.8
Armenian3OpenAI GPT-5.6 Terra254.3
Assamese3OpenAI GPT-5.6 Luna136
Azerbaijani1OpenAI GPT-5.574.4
Basque3OpenAI GPT-5.6 Sol208.8
Belarusian1Anthropic Claude Fable 595.7
Bengali1Anthropic Claude Fable 572
Bosnian2Anthropic Claude Fable 561103.1
Bulgarian1OpenAI GPT-5.6 Terra26
Catalan1OpenAI GPT-5.563.3
Cebuano3Google Gemini 3.1 Pro Preview293.1
Chinese Simplified3Anthropic Claude Fable 5166.2
Chinese Traditional3OpenAI GPT-5.6 Sol139.7
Croatian1Anthropic Claude Fable 560.6
Czech1OpenAI GPT-5.6 Terra27.7
Danish1OpenAI GPT-5.6 Terra24.7
Dutch1OpenAI GPT-5.6 Sol58.1
Estonian1OpenAI GPT-5.6 Sol78.8
Farsi1Google Gemini 3.1 Pro Preview68.1
Finnish1OpenAI GPT-5.6 Sol66.7
French3Anthropic Claude Fable 5172.7
Frisian3Google Gemini 3.1 Pro Preview280.9
Galician1OpenAI GPT-5.6 Sol47.7
Georgian2OpenAI GPT-5.5120.1259.2
German2OpenAI GPT-5.6 Luna30.279.3
Greek1Anthropic Claude Fable 594.5
Gujarati2Google Gemini 3.1 Pro Preview206.2
Hebrew1Anthropic Claude Fable 585.9
Hindi2Anthropic Claude Fable 565.8
Hungarian1Anthropic Claude Fable 574.3
Icelandic2OpenAI GPT-5.5206
Indonesian1Google Gemini 3.5 Flash11.9
Irish3OpenAI GPT-5.6 Luna138.5
Italian2OpenAI GPT-5.6 Luna12.1
Japanese3Anthropic Claude Fable 5191.4
Kannada3Google Gemini 3.1 Pro Preview282.5
Kazakh2OpenAI GPT-5.5108.2
Khmer1OpenAI GPT-5.5215.5
Korean3Google Gemini 3.5 Flash118.3
Kyrgyz3OpenAI GPT-5.6 Terra195.8
Lao2Google Gemini 3.1 Pro Preview162.5
Latvian2OpenAI GPT-5.568.4
Lithuanian2Google Gemini 3.1 Pro Preview99.2
Macedonian1OpenAI GPT-5.6 Terra31.7
Malay1OpenAI GPT-5.6 Luna12.7
Malayalam2Google Gemini 3.1 Pro Preview205.5
Marathi3Google Gemini 3.1 Pro Preview273.8
Mongolian3Google Gemini 3.1 Pro Preview325
Nepali1OpenAI GPT-5.5118.3
Norwegian1OpenAI GPT-5.6 Luna12.1
Odia (Oriya)3Google Gemini 3.1 Pro Preview371.8
Pashto3OpenAI GPT-5.5361
Polish2OpenAI GPT-5.6 Terra46.7
Portuguese (European)2OpenAI GPT-5.6 Luna11.5
Punjabi2OpenAI GPT-5.6 Sol257.6
Romanian1Anthropic Claude Fable 559.7
Russian1OpenAI GPT-5.6 Sol43.9
Serbian1Anthropic Claude Fable 562
Sinhala3OpenAI GPT-5.6 Luna135.8
Slovak1Anthropic Claude Fable 565.8
Slovenian1OpenAI GPT-5.554.9
Spanish (European)1OpenAI GPT-5.6 Terra17.4
Swahili3OpenAI GPT-5.6 Luna118.9
Swedish1OpenAI GPT-5.550.8
Tagalog3OpenAI GPT-5.6 Terra164.6
Tajik3OpenAI GPT-5.6 Sol271.4
Tamil2OpenAI GPT-5.6 Sol200.1
Tatar3Translated Lara 2254.5
Telugu2Anthropic Claude Fable 5175.5
Thai3Google Gemini 3.5 Flash164.7
Turkish1OpenAI GPT-5.545.4
Ukrainian1Google Gemini 3.1 Pro Preview82.6
Urdu2OpenAI GPT-5.5114.4
Uzbek3OpenAI GPT-5.5222.7
Vietnamese1Anthropic Claude Opus 4.866.8
Welsh1Anthropic Claude Fable 591.2
Yiddish3Anthropic Claude Fable 5407.6

Units — s per 1,000 words: s per 1,000 words; Fastest one-step solution, s per 1,000 words: s per 1,000 words

Steps: 1 for a one-step solution, 2 for a two-step solution (translation + post-edit, or translation + repair by code), 3 for a three-step solution (translation + LQA + repair). For three-step solutions the time includes the LQA step, an estimate. In the last column, the time of the fastest one-step solution where a two-step solution is faster.

Automate translation to meet your requirements

See Intento Language Hub in action.

Book a demo
Leave feedback