Google Launches Gemini 3.8 Live Models With Native Speech-to-Speech and Background Tool Calls

·1 views

Google announced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on September 15, 2026, marking a shift from traditional speech-to-text-to-speech pipelines toward native, continuous voice processing. The two models serve distinct purposes: the standard Live model prioritizes fluid, low-latency dialogue at scale, while the Extended Thinking variant handles complex multi-step reasoning and slow tool calls while narrating progress aloud. A key technical feature is asynchronous function calling, which allows background tasks to execute without interrupting the audio stream, reducing awkward silences in voice interactions. The models support over 97 languages and are being rolled out across Google surfaces including Gmail Live, Docs Live, and Search Live. Google has priced the Live API at $0.005 per minute for audio input and $0.018 per minute for output, making production-scale voice agent deployment more financially viable for product teams.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Kyber (YC W23) Seeks Forward Deployed Engineer

Kyber, a startup that participated in Y Combinator's Winter 2023 batch, is currently hiring for a Forward Deployed Engineer role. The job listing was posted on Y Combinator's company job board. A Forward Deployed Engineer typically works closely with customers to implement and support technical solutions. The posting received no points or comments on Hacker News at the time of publication.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

How One Dev Team Ditched WordPress to Unify 50+ Retail Sites on a Headless Stack

A retail group that grows by acquiring independent shops tasked a development team with maintaining five luxury-segment WordPress websites, each inherited with inconsistent plugins, themes, and product data. The team first stabilised the sites through updates and audits, then evaluated over a dozen centralised management tools, none of which solved the core problems of decentralised product data and slow site deployment. After weeks of failed attempts using custom CSV scripts and WP-CLI workflows, the team concluded that WordPress could not support the client's model of adding a new website with every acquisition. The chosen solution moved all product and content data into a single Strapi CMS instance, with individual static frontends built in Astro for each site. The client approved the new headless architecture, addressing both the central data management gap and the need to spin up new sites far more quickly.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Engineers Are Cutting AI Response Latency by 70% in Real-Time Apps

Developers working on mission-critical enterprise applications are adopting multi-tier strategies to reduce AI latency and improve real-time performance. Key techniques include semantic caching for repeated queries, streaming responses to lower perceived wait times, and routing urgent tasks to smaller, hardware-accelerated local models. For globally distributed systems, deploying AI inference at regional edge nodes helps minimize network round-trip delays. A ride-sharing platform reportedly applied edge caching and lightweight neural networks to slash dispatch calculation latency by 70% during peak hours. These architectural approaches are increasingly seen as essential for delivering seamless user experiences in high-demand AI systems.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

x402 Protocol Lets AI Agents Pay for API Calls Autonomously Without Human Input

The x402 HTTP payment protocol, built on the long-dormant HTTP 402 'Payment Required' status code from 1991, now enables AI agents to autonomously pay for API calls using stablecoins in a single round-trip exchange. When an agent receives a 402 response, it pays and retries the request automatically, eliminating the need for pre-registered API keys or human involvement in each transaction. Supporting this is WAIaaS, an open-source self-hosted Wallet-as-a-Service built specifically for AI agents, which separates spending authority across three roles: the agent, the administrator, and the fund owner. The system enforces a policy engine with 21 policy types and four security tiers — ranging from instant execution to mandatory human approval — keeping humans in control without making them a bottleneck for every payment. Developers can deploy this infrastructure today, addressing a key scaling challenge as multi-agent systems require real-time, programmatic financial decision-making.

0 comments Read more at DEV Community