Best Translation APIs Compared: GPT, Claude, DeepL & Google (2026)
A developer's buying guide to the 2026 translation API market. Covers the four kinds of translation API now on offer, a head-to-head scorecard, honest strength-and-failure profiles for each vendor, current rate cards by named model for GPT-5.x and Claude Opus 5 / Sonnet 5 / Haiku 4.5 alongside DeepL and Google Cloud, real cost math on a 2,000-string catalog across 12 languages (LLMs now undercut NMT on UI strings), how batch endpoints and prompt caching cut the bill by 80%+ and the two ways caching silently fails, schema-constrained output as the biggest reliability upgrade available, placeholder integrity per engine, the plural-forms cliff, latency and token-per-minute limits, data retention and compliance questions, a pick-by-scenario table, and the routing architecture most teams converge on.