No assessed real-time one-step solution meets every Standard requirement.
Real-time one-step solutions by model, and the Standard requirements they meet
| Model | Type | Languages with a real-time solution | Meet general translation quality | Meet every Standard requirement | Met in 50% of the languages or fewer |
|---|---|---|---|---|---|
| DeepL Next Gen | LLM-based MT engine | 71 | 2 | 0 |
|
| Translated Lara 2 | LLM-based MT engine | 38 | 0 | 0 |
|
| Microsoft Translator | MT engine | 32 | 0 | 0 |
|
| Google Translation LLM | LLM-based MT engine | 31 | 0 | 0 |
|
| OpenAI gpt-oss-120b | open-weight LLM | 10 | 0 | 0 |
|
| Amazon Translate | MT engine | 7 | 0 | 0 |
|
| ModernMT Enterprise Edition | MT engine | 5 | 0 | 0 |
|
| Google Cloud Advanced Translation API | MT engine | 4 | 0 | 0 |
|
Real-time: 2 seconds or less per text (~744 words on average).
The fastest one-step solution that meets every Standard requirement, by language
| Language | Model | s per 1,000 words | s per text |
|---|---|---|---|
| Indonesian | Google Gemini 3.5 Flash | 11.94 | 8.89 |
| Norwegian | OpenAI GPT-5.6 Luna | 12.14 | 9.04 |
| Malay | OpenAI GPT-5.6 Luna | 12.72 | 9.47 |
| Spanish (European) | OpenAI GPT-5.6 Terra | 17.45 | 12.99 |
| Danish | OpenAI GPT-5.6 Terra | 24.67 | 18.37 |
| Bulgarian | OpenAI GPT-5.6 Terra | 25.99 | 19.35 |
| Czech | OpenAI GPT-5.6 Terra | 27.7 | 20.62 |
| Macedonian | OpenAI GPT-5.6 Terra | 31.66 | 23.57 |
| Russian | OpenAI GPT-5.6 Sol | 43.94 | 32.71 |
| Turkish | OpenAI GPT-5.5 | 45.41 | 33.81 |
| Galician | OpenAI GPT-5.6 Sol | 47.71 | 35.52 |
| Swedish | OpenAI GPT-5.5 | 50.83 | 37.84 |
| Slovenian | OpenAI GPT-5.5 | 54.95 | 40.91 |
| Arabic (MSA) | Anthropic Claude Fable 5 | 57.8 | 43.03 |
| Dutch | OpenAI GPT-5.6 Sol | 58.07 | 43.23 |
| Romanian | Anthropic Claude Fable 5 | 59.68 | 44.43 |
| Croatian | Anthropic Claude Fable 5 | 60.6 | 45.12 |
| Serbian | Anthropic Claude Fable 5 | 62.04 | 46.19 |
| Catalan | OpenAI GPT-5.5 | 63.34 | 47.16 |
| Slovak | Anthropic Claude Fable 5 | 65.82 | 49 |
| Finnish | OpenAI GPT-5.6 Sol | 66.69 | 49.65 |
| Vietnamese | Anthropic Claude Opus 4.8 | 66.82 | 49.75 |
| Farsi | Google Gemini 3.1 Pro Preview | 68.09 | 50.69 |
| Bengali | Anthropic Claude Fable 5 | 71.98 | 53.59 |
| Hungarian | Anthropic Claude Fable 5 | 74.35 | 55.35 |
| Azerbaijani | OpenAI GPT-5.5 | 74.4 | 55.39 |
| Estonian | OpenAI GPT-5.6 Sol | 78.84 | 58.7 |
| German | Google Gemini 3.1 Pro Preview | 79.3 | 59.04 |
| Ukrainian | Google Gemini 3.1 Pro Preview | 82.61 | 61.5 |
| Albanian | Anthropic Claude Fable 5 | 85.18 | 63.42 |
| Hebrew | Anthropic Claude Fable 5 | 85.9 | 63.95 |
| Welsh | Anthropic Claude Fable 5 | 91.16 | 67.87 |
| Greek | Anthropic Claude Fable 5 | 94.51 | 70.36 |
| Belarusian | Anthropic Claude Fable 5 | 95.74 | 71.28 |
| Bosnian | Google Gemini 3.1 Pro Preview | 103.14 | 76.79 |
| Nepali | OpenAI GPT-5.5 | 118.31 | 88.08 |
| Khmer | OpenAI GPT-5.5 | 215.51 | 160.45 |
| Georgian | Anthropic Claude Fable 5 | 259.15 | 192.94 |
Units — s per 1,000 words: s per 1,000 words; s per text: s per text
The languages where at least one one-step solution meets every Standard requirement. The model is adjusted with tone of voice, terminology and other context in the request.