July 15, 2026
Lost in translation?
If you want Claude to speak nicely to you, try Hindi or Arabic
Claude gets sweeter in Hindi and Arabic — and the comments are absolutely not charmed
TLDR: Anthropic says Claude’s tone changes by language, sounding warmer in Hindi and Arabic and more strict in English. Commenters were far less polite: one said the company’s original post was the only interesting part, while another bluntly declared that Claude just isn’t very good.
Anthropic just dropped a fascinating claim: its chatbot Claude doesn’t sound the same in every language. In plain English, the bot comes off more warm and deferential in Hindi and Arabic, more strict and rigorous in English and Russian, more humble in Dutch, and more slick and confident in Indonesian. That’s the official finding. But in the comment section, the real drama was less “wow, language is complex” and more “hang on, did anyone even read the better version?”
One of the strongest reactions came from readers basically side-eyeing the article itself. User malshe shrugged off the write-up as a watered-down recap and pointed everyone to Anthropic’s own original post, calling it “much more interesting” and praising its visual charts. Ouch. That’s classic internet energy: nice article, but the source link is where the good stuff is. Then came the pure blunt-force take from verzali: “Claude kind of sucks, to be honest.” No nuance, no ceremony, just straight to the jugular.
And that’s the vibe in a nutshell. Anthropic is trying to map how its AI changes personality depending on language, while readers are split between mild curiosity, skepticism, and outright dismissal. The accidental comedy here is delicious: a story about whether Claude sounds nicer somehow inspired commenters to be spectacularly not nice about Claude itself.
Key Points
- •Anthropic researchers reported that Claude’s responses vary across languages along four axes: Deference vs. Caution, Warmth vs. Rigor, Depth vs. Brevity, and Candor vs. Execution.
- •Anthropic said these four axes account for 15 percent of the variation in the values or behavioral patterns Claude expresses across languages.
- •The company clarified that Claude does not intrinsically hold values; the term refers to values reflected in its behavior and outputs.
- •The article says Sonnet 4.6 tends to appear more deferential and warm, while Opus 4.7 appears more focused on accuracy, precision, and misuse prevention.
- •Anthropic’s cited findings associate Arabic and Hindi with warmer expression, English and Russian with greater rigor, Dutch with more candor, Indonesian with more polished confidence, Arabic with brevity, and English with more depth.