Is Deepseek v3.1 Terminus good at Translation & Localization?
Deepseek v3.1 Terminus ranks #43 of 44 for Translation & Localization — strong. The top pick for this task is qwen3.7-max.
Deepseek v3.1 Terminus on each Translation & Localization sub-task
| Register & Formality | 97.0/100 | #44 |
| Business Translation | 90.0/100 | #43 |
| Localization | 74.0/100 | #40 |
| Catch the Translation Error | 71.0/100 | #41 |
Real examples, graded
WeakSupport reply with a false-friend trap (EN→German) 70/100
“The model failed the specific instruction to handle 'embarrassed' carefully. It explicitly rejected the correct natural translations ('peinlich', 'unangenehm') and instead provided a translation that loses the nuance ('tut uns leid') and one that is too dramatic for business register ('schämen uns'). Additionally, the first sentence contains translationese ('Es tut uns leid für').”
WeakJapanese keigo direction (EN→Japanese) 91/100
“The model uses excellent, natural business Japanese with the correct humble keigo (kenjougo) for the company's actions. However, it transcreates the message rather than translating it directly, omitting 'your request' and adding 'we will contact you again'. It also includes unprompted explanatory text.”
Frequently asked
Is Deepseek v3.1 Terminus good at Translation & Localization?
Deepseek v3.1 Terminus ranks #43 of 44 models we tested for Translation & Localization, scoring strong.
What is Deepseek v3.1 Terminus's strongest Translation & Localization skill?
Its best sub-task here is Register & Formality.
This page is Spring Prompt, running
We just did this for every model. Do it for your prompt.
The rankings above come from running real tasks through real models and scoring every output. Spring Prompt is that same engine — pointed at your prompt, your test cases, and your definition of good.
- Generate test cases from your prompt — no eval set required to start.
- Compare models side by side with quality, cost and latency in one matrix.
- Optimise the winner until the scores say it's ready to ship.
Prompt × model results
12 test cases · 3 evals