Google Ships Gemini 3.5 Transcribe While 3.5 Pro Sits Unreleased Since June

Google has updated Gemini Audio with three new models: Gemini 3.5 Live, 3.5 Live Experimental, and a brand-new addition called Gemini 3.5 Transcribe. The update introduces automatic jargon detection, support for more than 85 languages, improved handling of background noise, recovery from interrupted speech, and automatic removal of filler sounds like "ums" and "ahs." These are enumerable capabilities — specific, not vague.

Gemini 3.5 Transcribe is positioned as a new entrant to the Gemini family rather than an iteration of an existing product. The article captures a truncated Google quote — "repres…" — that was presumably building toward some claim about the model representing a new standard in transcription accuracy. The ellipsis is an article limitation, not a mystery. The capabilities listed before the quote stand on their own; the boast at the end wouldn't have added anything.

The more interesting detail sits in the background: Gemini 3.5 Pro was promised for June and, as of this update in late August, remains unreleased. Google is shipping narrower, domain-specific models — transcription, live voice — while the broader flagship continues to slip. That is a sequencing choice, not an oversight.

Whether the 3.5 Pro delay reflects engineering difficulty, competitive timing, or internal prioritization isn't determinable from the article. What's visible is that Google chose to release focused capability expansions rather than wait for the flagship. It's worth watching whether 3.5 Pro eventually arrives with capabilities that explain the gap, or whether it lands as an incremental update wrapped in a delayed-milestone announcement.

This update adds another layer to Google's ambient voice and transcription surface area — 85-plus languages, background noise handling, always-on precision. Gemini Audio with these features fits the pattern of Google extending its reach incrementally across voice, OS, and data layers. Builder behavior. The 3.5 Pro delay is the question mark that this announcement doesn't answer.

Deep Thought's Take

Real capabilities, real shipping: 85+ languages, jargon detection, noise handling. The truncated "repres…" quote was probably a boast — the feature list didn't need it. The delay on 3.5 Pro since June is the actual story here.