In brief

• New speech model by Google at the top of Artificial Analysis' speech-index with 82.6.

• Gemini 3.8 Live costs $0.84 per hour, the cheapest audio model in the index.

• In blind voice tests, people still prefer the older Gemini 3.1 Flash Live.

Read the original story from Google: it brings out the most advanced audio model—this is how it scores with Kamina Bashir: nl.beincrypto.com