What happened
Google Deepmind released Gemini 3.8 Live and 3.8 Live Extended Thinking via the Gemini API and Google AI Studio. Extended Thinking ranks first on the Artificial Analysis Speech-to-Speech Leaderboard with 82.6 percent, ahead of OpenAI's GPT-Live-1; Google charges $0.005 per minute for audio input and $0.018 for output.
Why it matters
The top leaderboard ranking puts Google's audio model ahead of OpenAI's latest voice model, and the per-minute input and output prices undercut OpenAI's — signaling Google is competing on cost as well as capability.
What to watch
OpenAI's model should still deliver more natural conversations thanks to full duplex, and judging from the demos it sounds better, suggesting Google may have optimized for price over quality.
WHO IT HITSDevelopers building voice agents on the Gemini API or Google AI Studio now get a cheaper, leaderboard-topping audio option, though those prioritizing natural-sounding conversation may still lean toward OpenAI's GPT-Live-1.
Summaries like this, every morning.
Google Deepmind's release of Gemini 3.8 Live and 3.8 Live Extended Thinking marks a direct push into voice agents that developers can build on through the Gemini API and Google AI Studio. The Extended Thinking variant's first-place finish on the Artificial Analysis Speech-to-Speech Leaderboard with 82.6 percent, ahead of OpenAI's latest GPT-Live-1 models, gives Google a measurable claim to quality leadership in this category. Google also supports over 97 languages and points developers to sample apps on GitHub, lowering the barrier to trying the models.
The pricing gap is the other half of the story. Google charges $0.005 per minute for audio input and $0.018 for output, while OpenAI's GPT-Live-1 sits at $0.05 per minute. Once those rates are applied to a full hour of voice conversation, Google costs about $1.38 versus at least $3.00 for OpenAI.
What remains open is whether Google's advantage holds up in practice. OpenAI's model should still deliver more natural conversations because of full duplex, which lets it listen and speak at the same time, and the demos suggest it sounds better — so Google may have traded some conversational polish for price. The leaderboard score and the per-minute rates are the concrete pieces to weigh; how each model actually feels in live back-and-forth is likely to decide which one developers pick.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Amphenol reported adjusted earnings of 80 cents per share, up 74% year over year, and revenues of $10.23 billi…
Analog Devices, collaborating with Nvidia on AI data-center infrastructure, physical AI and edge robotics, ran…
Charles Schwab is expanding AI tools for registered investment advisors, integrating Claude for Financial Advi…
Octus says it routes more than 20 billion tokens a year through its data science models, using what it calls a…
T-Mobile beat forecasts and set out a mid-term plan covering an AI build-out and a pay-out
Morningstar published a piece calling Amphenol a well-diversified infrastructure behemoth riding the AI tidal…
