An AI translation platform by Tomedes that runs text and documents through 22 AI models in parallel and selects the version most models agree on (branded SMART), rather than relying on a single engine. It supports 330+ languages and 72,600+ language pairs, preserves document layout (PDF, DOCX, XLSX, and images), and provides side-by-side comparison with quality scores. Pricing: Free (60 words per translation, unlimited daily), Pro **$19/mo** (promotional, standard $39/mo, unlimited text/documents), pay-per-translation **$2.50** (or $1.50 for ≤300 words), short-term passes ($6/1 day, $9/3 days, $12/5 days), and human verification at $0.04/word (2,000 free words/month on Pro). An MCP server is available on paid plans for integration with Claude, ChatGPT, and Cursor. Caveat: a Trustpilot reviewer reported that a paid “unlimited” monthly plan stopped working after 5 days with no customer service response. G2 reviews (131 total) are largely positive, praising multi-engine comparison and format preservation.
Pros & Cons:
Pros: The multi-model consensus approach is a meaningful alternative to single-engine translators—seeing 22 outputs and the version most agree on reduces the risk of a single model’s idiosyncratic error, which is useful for professional or high-stakes content; the side-by-side comparison and quality scores give users visibility into why one version might be better than another, rather than a black-box result; document translation preserves layout across common formats, and the MCP server (paid plans) lets you translate inside Claude/ChatGPT/Cursor without switching tabs; the free tier is usable for short texts (60 words per translation, unlimited times), and the Pro plan includes 2,000 words of human verification per month.
Cons: The “unlimited” Pro plan has a documented reliability complaint—a Trustpilot reviewer reported it stopped working after 5 days with no support response, which is concerning for anyone relying on it for ongoing work; the free tier is heavily limited to 60 words per translation (longer texts show only a preview), making it unsuitable for evaluating document translation quality; quality depends on the models in the consensus pool, and the platform’s own blog acknowledges that lower-tier models (available to free users) produce quieter, subtler errors on complex text compared with premium models (paid users), so free-tier output is not representative of paid quality; the MCP server and API are paid-only, and MCP usage consumes credits per call rather than per word, which can be harder to budget; it is not a replacement for professional human translation for legal, medical, or regulatory content—human verification is an add-on, not a guarantee.
Who is it for?:Professionals who want to compare multiple AI translation outputs and choose the best one, rather than trusting a single engine; teams translating documents (PDF, DOCX, XLSX) that need layout preservation; developers who want translation accessible inside Claude/ChatGPT/Cursor via MCP (paid plans); users translating short texts frequently but not needing unlimited volume (free tier or pay-per-translation). Not suited for users who want a single, best-in-class translation engine (DeepL is more focused on quality), anyone relying on the “unlimited” Pro plan for critical ongoing work without verifying reliability, or those needing free document translation (free tier is text-only, 60 words).