DeepL Review (2026): Pricing, Free Tier, and Translation Test

👁 121 reads

DeepL is a Cologne-based AI translator supporting 114 languages, offering a free tier across web and desktop apps, with the paid Individual plan starting at 8.99 EUR per month (approx. $10.05 USD) or 7.49 EUR per month (approx. $8.37 USD) billed annually (conversions based on the ECB reference rate of 1.1181 dated 2026-10-12 via ecb.europa.eu; estimates, not verified US checkout prices). Its companion editing tool, DeepL Write, supports only 11 languages and variants. In our translation test of three idiom-dense paragraphs translated into English, modern LLMs came out on top (GPT-6.1 Sol scored 20 out of 20 points), while Google Translate translated 7 out of 14 tricky spots literally. A broader test covering 70 sentences across 8 translators is detailed in our guide to the best AI translator.

What is DeepL?

DeepL is a German company founded in Cologne in 2017 by the team behind the Linguee dictionary, led since its inception by Jaroslaw Kutylowski. It originated as a browser-based translation tool celebrated for sounding substantially more natural than Google Translate. Today, it operates as a multi-product platform:

• DeepL Translator - Translates text, files (Word, PowerPoint, Excel, PDF), text from images, and complete web pages.
• DeepL Write - Text editing and rewriting, currently limited to 11 languages and variants (English, German, French, Spanish, Italian, Portuguese, Japanese, Korean, and Chinese).
• DeepL Voice - Speech translation for meetings and real-time dialogue, including 30 minutes per user per month on paid plans.
• DeepL API - Automated programmatic translation with glossary support and translation memory integration.

DeepL runs in standard web browsers, via desktop apps for Windows and Mac, mobile apps for Android, iPhone, and iPad, browser extensions for Chrome, Edge, and Firefox, and add-ins for Microsoft Word, Outlook, PowerPoint, and Google Workspace. The company was valued at $2 billion in 2024.

How to get started: step-by-step translation

1. Go to deepl.com/translator or install the desktop application. An account is not strictly required for quick translations, but registering a free account increases limits and unlocks glossary features.
2. Paste text or drag and drop a document. The source language detects automatically; select the target language from the list.
3. Click on any translated word to inspect alternate phrasings. Choosing an alternative dynamically rephrases the surrounding sentence.
4. Add persistent brand or technical terminology to your glossary (the free plan includes one glossary with 5 entries per language pair).
5. On the desktop client, use global keyboard shortcuts like Ctrl+C+C: highlight text in any program, tap C twice while holding Ctrl, and the translation opens instantly in an overlay.

For more detailed strategies on text preparation, glossaries, and quality control, explore our guides on AI translation tools and how to translate text while preserving meaning.

DeepL pricing and plans

DeepL’s European pricing is listed in euros with VAT included, checked on October 3, 2026. Approximate USD equivalents are calculated using the European Central Bank reference rate of 1.1181 dated 2026-10-12 (ecb.europa.eu):

All paid plans include explicit contractual guarantees that user texts are never stored or used to train models, alongside a 30-day free trial (excluding Business). Write Pro is an optional add-on priced at 6.99 EUR per month (~$7.82 USD) billed annually or 8.99 EUR monthly (~$10.05 USD). For a deeper breakdown of file allowances and team permissions, see our guide on DeepL Pro pricing.

DeepL Free: what is included?

• Text translation with a cap of 50,000 characters per month.
• One document translation per month (up to 5 MB).
• One glossary with up to 5 custom entries per language pair.
• Access to official desktop apps and browser extensions.

The primary compromise is data privacy: on the free tier, DeepL reserves the right to use submitted text to improve its underlying translation models. Confidential contracts, client records, and sensitive corporate files should never be pasted into any free translation tool.

Translation benchmark: DeepL rivals and LLMs tested

We tested translation into English using a deliberately challenging source passage: three paragraphs (143 words) packed with 13 idioms, gross and net financial figures, comma-separated decimals, a date in European format (15.11.2026), and specific times. We tested it in Polish, feeding the identical text on October 3, 2026 to Google Translate, four frontier language models via OpenRouter, and a locally hosted 11B model (Bielik 11B) running on an RTX 4090 GPU.

Transparency note on DeepL: The free DeepL API requires an active registration key (Developer tier), while web-interface testing was omitted to ensure every benchmarked tool processed the exact source text uniformly via automated API pipelines without manual clipboard intervention. DeepL is omitted from this programmatic benchmark table. Google Translate and the frontier LLMs serve as baseline comparisons, and the source text can be tested independently against DeepL’s web interface using our test files linked below.

* The host GPU was concurrently processing other background workloads, inflating latency for the local model.

Key takeaways from the benchmark:

• Numeric formatting is solved. Every tested engine correctly converted currency notations, decimal points (“3,5%” to “3.5%”), and dates into standard English format.
• Google Translate still struggles with idiomatic nuance: It translated idiomatic phrases word-for-word into nonsensical literal phrases (e.g., rendering Polish idioms as “pears on willow”, “mustard after dinner”, or “splitting the skin of a bear”). A human resources manager (“kadrowa”) was mistranslated as “the staff”, and the narrator’s grandmother was misattributed as “his grandmother”.
• LLMs handle cultural equivalents accurately: Abstract idioms were naturally adapted into standard English equivalents such as “promised the moon”, “pie in the sky”, or “too little, too late”.
• Subtle semantic shifts: Claude and Gemini rendered an idiom meaning “breaking even” as “come out ahead” (implying net profit). DeepSeek translated an idiom meaning “not his cup of tea” literally as “it wasn’t his job”.

For an itemized breakdown of all 20 evaluation checkpoints, side-by-side excerpts, and architectural differences, review our complete comparison in DeepL vs Google Translate vs ChatGPT.

DeepL compared to alternatives

DeepL holds a distinct advantage in document workflows: it converts slide decks, contracts, and spreadsheets while maintaining strict visual layouts and glossary adherence. General-purpose LLMs excel with humor, idiomatic dialogue, and specialized tone requirements because you can instruct them directly on context and target readership. Google Translate remains the leader in raw language coverage and speed, making it suitable for quick gist comprehension. For mobile document scanning, refer to our walkthrough on photo translation tools.

How to get better translation quality from DeepL

Most errors encountered in automated translation stem from issues in the source material rather than the translation engine itself. Before submitting text to DeepL, follow these core principles:

• Break down complex sentences. Sentences running four or five clauses with ambiguous relative pronouns are the primary cause of mistranslated subjects and misplaced modifiers.
• Clarify idioms in functional text. In technical manuals or contracts where precision is critical, replace ambiguous figures of speech with literal descriptions before running the translation.
• Populate glossaries upfront. Establish glossary terms before processing large document batches rather than editing outputs retroactively.
• Select tone explicitly. Configure the formal or informal register toggle for target languages that require grammatical distinction (such as German “Sie” versus “du”).
• Run back-translations for critical passages. Translate key contractual clauses back into your source language to verify whether the original intent and legal scope remain intact.

DeepL integration with ChatGPT, Claude, and Copilot

With DeepL’s Model Context Protocol (MCP) server, users can connect DeepL directly as an execution tool inside Microsoft Copilot, ChatGPT, and Claude. In this hybrid setup, the conversational LLM orchestrates context and formatting, while DeepL executes the actual translation adhering to your locked corporate glossaries. This provides an effective setup for enterprises maintaining strict glossaries inside DeepL while operating primarily within generative AI interfaces.

Who should use DeepL?

• Recommended for: Professionals translating structured documents (proposals, technical manuals, slide decks) who require preserved page layouts; organizations needing strict zero-retention data privacy agreements; teams managing standardized enterprise terminology.
• Less suited for: Advanced grammar correction in unsupported languages (DeepL Write is restricted to 11 options; see our comparison of Grammarly, QuillBot, and DeepL Write), highly creative wordplay, or low-resource dialects not included in its 114 supported languages.
• Plan selection: Free for infrequent quick checks, Individual for routine file translation, and Team when shared terminology glossaries are required across multiple seats.

DeepL is also expanding into voice translation infrastructure. For more details on their acquisition of Mixhalo and live multilingual event support, see our coverage on DeepL’s acquisition of Mixhalo.

Frequently asked questions

Is DeepL completely free?

Yes, DeepL provides a permanent Free plan with a cap of 50,000 characters per month, one document translation up to 5 MB per month, and one glossary. Paid Pro subscriptions start from 7.49 EUR per month (~$8.37 USD) billed annually.

How much does DeepL Pro cost?

DeepL’s European pricing is listed in EUR inclusive of VAT: Individual costs 8.99 EUR monthly (~$10.05 USD) or 7.49 EUR monthly (~$8.37 USD) when billed annually. Team tiers cost 29.99 EUR monthly (~$33.53 USD) or 24.99 EUR monthly (~$27.94 USD) per user billed annually. Official USD checkout prices may vary depending on local tax requirements and checkout billing regions.

DeepL vs Google Translate: which is better?

DeepL and modern LLMs deliver more natural phrasing and significantly better handling of idioms and tone. In our benchmark, Google Translate rendered 7 out of 14 difficult idiomatic phrases literally, though it handled numbers and calendar dates accurately. Google Translate still leads in language variety, supporting more than 240 languages compared to DeepL’s 114.

Does DeepL Write support all languages?

No. DeepL Write is currently restricted to 11 languages and regional variants, including English, German, French, and Spanish. For grammar editing across other languages, tools like LanguageTool or targeted prompts in frontier LLMs are more suitable.

Is DeepL secure for confidential business documents?

On paid Pro plans, DeepL provides contractual guarantees that submitted text is never used to train its models and is deleted immediately after translation. The Business tier also includes enterprise security features like SSO, domain management, and SCIM provisioning. The Free tier does not include these data privacy guarantees.

Does DeepL offer desktop applications?

Yes. DeepL offers dedicated desktop clients for Windows and Mac, mobile apps for iOS and Android, extensions for Chrome, Firefox, and Edge, and workplace integrations for Microsoft 365 (Word, Outlook, PowerPoint) and Google Workspace.

Can DeepL translate PDF documents?

Yes. The Free plan permits 1 PDF up to 5 MB per month, the Individual plan allows 3 files up to 30 MB, and the Business plan allows up to 100 files up to 100 MB. Document formatting is typically preserved, although scanned PDFs without an embedded OCR text layer will yield lower quality.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Sources and test notes

Data verified and tested on October 3, 2026. The full benchmark dataset, prompt instructions, and raw translation outputs are documented in our linked comparison analysis.

• DeepL Pro pricing overview
• DeepL Translator supported languages list
• DeepL Write supported languages
• DeepL API language capabilities documentation
• DeepL press release: DeepL Voice speech preservation (September 15, 2026)
• OpenRouter model catalog

DeepL Voice for meetings

DeepL says its Voice models preserve elements of a speaker’s voice, including tone, pacing and intonation, across 14 supported languages. Voice-to-voice translation is available across more than 30 languages, so the broader translation coverage should not be confused with the narrower voice-preservation coverage. The company says voice-to-voice translation is generally available for online meetings, in-person conversations and through its API.

A separate Voice desktop app is available for Windows and Mac. It works alongside Zoom, Microsoft Teams and Google Meet, detects calls and provides an overlay for live translated subtitles. Users can switch from captions to voice-to-voice translation and adjust caption size, background audio levels and translation settings. See DeepL’s announcement for details.

Does DeepL Voice preserve voices in every supported language?

DeepL reports voice preservation in 14 languages, compared with more than 30 for voice-to-voice translation.

Which meeting platforms work with the Voice desktop app?

DeepL lists Zoom, Microsoft Teams and Google Meet.

Sources for this update

• https://www.deepl.com/en/press-release/deepl-voice-now-preserves-your-voice-in-real-time-multilingual-conversations

Kinetyra