LIVE
NVIDIA's Vera Rubin NVL72 Debuts in MLPerf Inference v6.116/09/26 · NVIDIA|OpenAI expands ChatGPT advertising with Sponsored Agents16/09/26 · OpenAI|OpenAI moves into advertising with 'Sponsored Agents'16/09/26 · OpenAI|Google DeepMind Introduces Gemini 3.8 Live and Its Extended Thinking Variant15/09/26 · Google DeepMind|What's at stake in AI's trillion-dollar infrastructure bet15/09/26|A Flaw in Chain-of-Thought Safety Monitoring14/09/26|Stellar Colosseum: A Multi-Agent System for Long-Horizon Mathematical Research14/09/26|Apple Code Hints Siri Could Be Swapped for ChatGPT or Claude14/09/26 · Apple|Anthropic says Houthi-linked actors used Claude Code for missile guidance software13/09/26 · Anthropic|Yoshua Bengio examines why AI agents lie, cheat and coordinate13/09/26|Terence Tao warns of a 'severe misalignment' in OpenAI's mathematical claims11/09/26 · OpenAI|Terence Tao warns of a 'severe misalignment' of AI in mathematics11/09/26|NVIDIA's Vera Rubin NVL72 Debuts in MLPerf Inference v6.116/09/26 · NVIDIA|OpenAI expands ChatGPT advertising with Sponsored Agents16/09/26 · OpenAI|OpenAI moves into advertising with 'Sponsored Agents'16/09/26 · OpenAI|Google DeepMind Introduces Gemini 3.8 Live and Its Extended Thinking Variant15/09/26 · Google DeepMind|What's at stake in AI's trillion-dollar infrastructure bet15/09/26|A Flaw in Chain-of-Thought Safety Monitoring14/09/26|Stellar Colosseum: A Multi-Agent System for Long-Horizon Mathematical Research14/09/26|Apple Code Hints Siri Could Be Swapped for ChatGPT or Claude14/09/26 · Apple|Anthropic says Houthi-linked actors used Claude Code for missile guidance software13/09/26 · Anthropic|Yoshua Bengio examines why AI agents lie, cheat and coordinate13/09/26|Terence Tao warns of a 'severe misalignment' in OpenAI's mathematical claims11/09/26 · OpenAI|Terence Tao warns of a 'severe misalignment' of AI in mathematics11/09/26|
ReleaseGoogle

Google adds Gemini 3.8 Live and an extended-thinking variant

Google has expanded its Gemini lineup with two new variants built for real-time voice and multimodal interaction, including a version designed to reason more extensively on complex requests.

September 15, 20263 min readPublished byHacker News

Google has announced on its official blog the arrival of Gemini 3.8 Live along with a variant called Gemini 3.8 Live Extended Thinking, two new additions to its model lineup built for real-time voice and multimodal interaction. These releases extend Google's existing Live API, already used to power conversational assistants capable of low-latency, fluid exchanges combining voice, image and text within a single interaction stream.

The Extended Thinking variant is set apart by an added deliberative reasoning step before generating a response, an approach Google had previously tested with 'thinking' variants of earlier Gemini generations. The stated goal is to help the model handle more complex or ambiguous queries without sacrificing the responsiveness expected from a real-time conversational system, a balance that remains technically difficult to strike.

The blog post itself is fairly light on technical specifics: benchmarks, pricing, regional availability and latency figures are not yet detailed in the initial announcement. The release fits into an increasingly rapid cadence of updates for the Gemini family, as Google appears intent on keeping pressure on rivals in the voice assistant and conversational agent space.

Without comparative performance data, it is difficult to gauge how much this version actually improves over prior iterations. Developers and companies relying on the Live API will likely need to wait for fuller documentation, or independent testing, to determine whether the added reasoning capability translates into meaningful gains for use cases such as enterprise voice agents or consumer-facing multimodal assistants.

Tags
geminigooglelive-apivoice-aimultimodalreasoning

Read also