Google releases Gemini 3.1 Flash Live, its highest-quality audio model for real-time voice AI
Google has released Gemini 3.1 Flash Live, its highest-quality audio and voice model designed for real-time dialogue. The model scores 90.8% on ComplexFuncBench Audio and 36.1% on Scale AI's Audio MultiChallenge with reasoning enabled, with improved tonal understanding and lower latency compared to previous versions.
Gemini 3.1 Flash Live — Quick Specs
Google Releases Gemini 3.1 Flash Live, Its Highest-Quality Audio Model
Google has launched Gemini 3.1 Flash Live, a real-time audio and voice model designed to deliver more natural and reliable voice interactions. The model is now available to developers via the Gemini Live API in Google AI Studio, to enterprises through Gemini Enterprise for Customer Experience, and to all users via Gemini Live and Search Live.
Performance Benchmarks
On ComplexFuncBench Audio—which measures multi-step function calling with various constraints—Gemini 3.1 Flash Live achieves 90.8%, outperforming the previous model. On Scale AI's Audio MultiChallenge, which tests complex instruction following and real-world audio conditions including interruptions and hesitations, the model scores 36.1% with "thinking" mode enabled.
Google claims the model delivers improved latency compared to its predecessor, enabling faster response times for voice-first applications. The company also reports enhanced tonal understanding, allowing the model to recognize acoustic nuances like pitch and pace, and to dynamically adjust responses based on user expressions of frustration or confusion.
Developer Features
For developers, Gemini 3.1 Flash Live enables building voice agents capable of executing complex, multi-step tasks in noisy environments. The model supports function calling with improved reliability at scale. In Gemini Live, users can maintain conversation context for twice as long as with the previous model, preserving continuity during extended brainstorming sessions.
Companies including Verizon, LiveKit, and The Home Depot have provided positive feedback on the model's performance in production workflows, highlighting natural conversation quality.
Multilingual and Global Rollout
Gemini 3.1 Flash Live is inherently multilingual, enabling this week's global expansion of Search Live to over 200 countries and territories. Users can now conduct real-time, multimodal conversations with Google Search in their preferred language.
Safety and Watermarking
All audio generated by Gemini 3.1 Flash Live is watermarked using Google's SynthID technology. According to Google, this imperceptible watermark is embedded directly into audio output, enabling reliable detection of AI-generated content to help prevent misinformation.
What This Means
Gemini 3.1 Flash Live represents a meaningful advancement in real-time voice AI, with concrete benchmark improvements in function calling and instruction following. The model's expansion to 200+ countries positions Google to compete more aggressively in voice-first AI interfaces. The SynthID watermarking approach addresses growing regulatory and safety concerns around synthetic audio detection. For enterprises and developers, the improved tonal understanding and lower latency reduce friction in deploying voice agents for customer service and complex task automation.
Related Articles
Mistral AI Releases Shieldstral-1.0-3B, a 3B-Parameter Policy-Adaptive Safety Classifier
Mistral AI has released Shieldstral-1.0-3B, a compact open-weight safety classifier that evaluates text and images against natural-language policies specified at inference time. The 3B model runs on a single GPU and reports F1 scores competitive with or exceeding larger moderation models like LlamaGuard-4-12B and GPT-OSS-Safeguard-20B on multiple benchmarks.
Mistral Releases Shieldstral, a 3B Open-Weights Safety Classifier That Matches Models 7x Its Size
Mistral has released Shieldstral, a 3B open-weights safety classifier that reframes content moderation as a policy-adaptive question-answering task. The model claims to match or outperform guard models up to 7x its size on text safety and multimodal benchmarks, and runs on a single 16GB GPU.
OpenAI Halts Parts of Astra Model Development After It Hit 'Critical' Cybersecurity Threshold
OpenAI disclosed that its in-development Astra model showed cyberattack capabilities strong enough that it cannot rule out a 'Critical' risk classification. The company has paused related internal activity and added security controls under its Preparedness Framework.
Mistral's 3B-Parameter Shieldstral Matches 20B Safety Model on Text Benchmarks
Mistral's new Shieldstral, a 3-billion-parameter open-weight safety classifier, posts an 84.9% F1 score on text benchmarks—tying OpenAI's GPT-OSS-Safeguard-20B, a model roughly seven times larger. The model lets operators define safety rules at runtime using plain-language yes/no questions instead of fixed taxonomies.
Comments
Loading...