AI Translation API Comparison

Choosing a translation provider is a multi-axis decision: price per character, language coverage, whether you need a managed glossary, whether you translate whole documents or just strings, whether a custom or adaptive model matters for your domain, and the commercial-use and data-retention terms. This page lays all of that out for the 10 options that developers and localization teams most often shortlist, drawn from 8 vendors, and ranks them by what they would actually cost at your volume.

1. The cost formula (dedicated MT)

DeepL, Google, Azure, Amazon Translate and ModernMT bill per character of source text. The tool normalises your volume to characters and subtracts any standing free tier:

monthly_cost = max(0, characters − free_tier) ÷ 1,000,000 × usd_per_million

If you enter words instead of characters, they are converted first: one word ≈ 6 characters (≈5 letters plus a space), the convention several MT vendors bill on. The data module cross-checks every dedicated-MT figure a second way — via the per-1,000-character rate — so the two routes must agree to the millionth of a dollar before the page will build.

2. LLM-as-translator (per token)

OpenAI, Anthropic Claude and DeepSeek bill per token, for both the source you send (input) and the translation they return (output). To compare them on the same per-character axis, the tool converts using two documented assumptions: about 4characters per token (OpenAI's published rule of thumb for English-like text) and an output roughly the same length as the input. That gives:

usd_per_million_chars = (input_price + output_price per 1M tokens) ÷ 4

These rows are marked est. The cross-check recomputes them the long way — via the raw token prices — and the two must match. Important caveat: Sinhala, Tamil and other non-Latin scripts tokenize into many more tokens per character than English, so an LLM's real cost on those languages is materially higher than the English estimate shown. For an exact token count, use the AI Token Counter.

3. Free tiers

Three providers here have a standing monthly free allowance, subtracted before billing: Azure Translator's F0 tier gives 2,000,000 free characters a month (the largest here), and both Google Cloud and DeepL Free give 500,000. DeepL Free's allowance is also a hard cap — beyond it you must upgrade to Pro, so the tool flags “free cap exceeded” rather than pretending the cost stays zero. Amazon Translate's 2M-character free tier is new-account-only for 12 months, so it is treated as zero. At small volumes a free tier can make a pricier per-character rate the cheapest overall — the ranking accounts for that automatically.

4. Quality is benchmark-dependent

The quality label on each provider paraphrases its vendor- or benchmark-published standing (WMT shared tasks, Artificial Analysis) — never our own measurement. Translation accuracy depends heavily on the specific language pair and the kind of text, so a model that leads on English↔German may trail on English↔Sinhala. Use the labels to shortlist two or three providers, then translate a sample of your own content on each before you commit.

5. Best-for badges

The “Cheapest at your volume”, “Lowest list rate”, “Most languages” and “Best free tier” callouts are derived deterministically from your current selection and volume. Cheapest is the lowest projected monthly cost (a free-only tier over its cap can't win), lowest-rate is the lowest published per-million-character price, most-languages is the highest documented language count, and best-free-tier is the largest standing monthly free allowance. Requiring a feature greys out providers that lack it without deleting them, so the comparison stays honest.

• DeepL — API pricing (Free and Pro)
• DeepL — developer docs (languages, glossary, documents)
• Google Cloud — Translation pricing (Basic v2, Advanced v3)
• OpenAI — API pricing (GPT-4.1 mini)
• Anthropic — Claude API pricing
• Microsoft Azure — AI Translator pricing
• Amazon — Translate pricing
• DeepSeek — API pricing
• ModernMT / Lara — pricing
• WMT24 — machine-translation shared-task results
• Artificial Analysis — model benchmarks

Every rate and capability flag was last cross-checked against these sources on 2026-06-28. Translation pricing, languages and models change frequently; this page is reviewed manually and whenever a provider announces a substantive pricing or model update. Quality figures are vendor/benchmark published and benchmark-dependent.