Machine Translation Engines · Microsoft

Microsoft Translator

Azure's translation API with the largest free tier in its class.

★ Top Pick — Machine Translation Engines. Highest measured score in this category: 98 / 100 (Grade A) in the Traditional Chinese Quality Benchmark v1.0, run 2026-11-21. See the run evidence →
98
Grade A · / 100
Traditional Chinese
Quality Benchmark v1.0
Script purity (simplified-character residue)99.0 / 100
Wording (mainland-term density)94.8 / 100
Terminology (preferred-term consistency)100.0 / 100

Measured 2026-11-21 · Browser-route supplement to Run 2026-11-21 (00:52-00:57 CST): bing.com/translator in a real browser session, free, not logged in, target set to Traditional Chinese, tone preset left at its default 'casual' setting (disclosed - see watch-outs). Earlier API/Edge channels were unreachable (archived). Raw outputs archived in the run evidence. Raw outputs are published in the benchmark run archive. Method: how we test.

Conditions disclosure: AI/LLM engines received the corpus with one shared neutral prompt ('Translate into Traditional Chinese as used in Taiwan' plus fixed formatting instructions). Machine-translation engines received the source text directly, with no prompt. The groups were measured under different conditions; this disclosure travels with the scores wherever they appear.

Microsoft Translator (Azure Translator in Foundry Tools) is Microsoft's machine translation API: text and document translation, language detection, transliteration and custom models trained on your own parallel data.

Strengths — from evidence & published facts

  • Perfect terminology record: all 35 checkpoints used the Taiwan-preferred rendering (35:0) - the strongest terminology result in the run
  • Only two wording flags, both documented substring boundary cases inside longer correct words (M1, S3)

Watch-outs

  • Condition difference: the web translator's tone preset was left at default 'casual', so the output uses informal address and colloquial phrasing throughout; the benchmark's three metrics do not score register or tone - judge formality separately
  • One simplified-residue character (G5); score is for the free Bing web translator, not the Azure API

Best for

Microsoft-stack organizations and API users who want a large free monthly allowance before paying per character.

Pricing (checked 2026-11-21)

Per-character API pricing with a generous free tier; commitment tiers lower the per-million rate at volume. Source: the vendor's official pricing page, checked 2026-11-21.

PlanPriceNotes
F0 — Free$0Standard + custom translation training, 2M characters/month
S1 — Pay as you go$10 per million charactersText translation
S1 — Document translation$15 per million charactersPer million characters
S1 — Custom translation$40 per million charactersPer million characters

Free tier: F0 free tier: 2 million characters/month free · Trial: F0 free tier (2M characters/month)

Visit the official site → Official pricing page

Score history (append-only)

Run dateBenchmarkScoreGrade
2026-11-21Traditional Chinese Quality Benchmark v1.098A

History is never rewritten: when a tool is re-tested, the new run is added and the old one stays.

Related in Machine Translation Engines

  • DeepL — Machine translation service with a strong reputation in European languages, plus an API.
  • Google Translate — The default machine translation engine — free on the web, per-character pricing on the API.
  • MyMemory — Translation-memory API with a free daily quota, by Translated.
  • Amazon Translate — AWS neural machine translation service with real-time and batch text/document translation APIs.
  • Yandex Translate — Yandex Cloud machine translation API, plus a consumer website for text, image and document translation.

Need this done by a human instead?

Tools route around quality problems; specialists remove them. HankLoc localizes websites, marketing, subtitles and games into Traditional Chinese — with a free one-page trial.

Localization services → Get a quote

Get the full benchmark report — and every future score — by email

The public pages show the headline scores. The email edition carries the complete dataset behind them:

  • The full per-segment benchmark dataset for every run (all 24 segments, every engine, every flag)
  • New-tool test alerts — the day a pending tool gets its first measured score
  • Quarterly re-test results, including score changes and what caused them

Get the benchmark report →

Free. One list, no spam, unsubscribe anytime. You will receive a confirmation email first.