🦢 TL;DR: nails the timing of listening, speaking and interrupting, but task completion still lags. Tencent unveils an omni interaction model unifying speech, video and agentic execution. Title: Omni Interaction Agent Technical Report (Gander) URL: 📌 Highlights 🧠 Cerebellum-Brain design: real-time control layer + Claude/Codex-powered reasoning layer 🎙 Streams audio/video in 1-second chunks, no external VAD needed ✅ 100% turn-taking accuracy vs GPT-Realtime's 96% ⚠️ Task completion (Pass@1) only 0.400, below GPT-Realtime's 0.600 🔀 Feeding transcribed text straight to the Brain lifts Pass@1 to 0.520 🌐 Fusing audio+video boosts Daily-Omni by +19.13pt over single-modality 📦 Weights, training code and eval harness planned for release The honest takeaway: nailing conversational timing and nailing task accuracy are still two separate problems. #MultimodalAI# #VoiceAgents#

Voice agents can handle increasingly complex calls, but often fail on the simple details. τ-Entity is a new benchmark measuring whether an agent can correctly capture details like names, phone numbers and email addresses, recognize when it might be wrong, and repair the mistake. Read more:

Voice agents are becoming a bigger part of customer and operational workflows globally, but getting them into production requires more than making the conversation sound natural. Join us for a practical session focused on a practical framework for evaluating production voice agents. We’ll cover how to assess tool use and task completion, connect conversations to real-world outcomes, and identify voice-specific issues like latency, interruptions, pacing, and conversational friction.

Voice Cloning is now live via the xAI API! Create a custom voice in less than 2 minutes or select from our library of 80+ voices across 28 languages to personalize your voice agents, audiobooks, video game characters, and more.

AI voice agents can complete common phone scams in about 130 seconds. They succeed 36% of the time, at under USD 0.75 per successful scam. TRM's new report measures how AI is changing the economics of fraud across four dimensions — speed, scale, cost, and expertise — and provides recommendations for closing the capacity gap. Read the full report here ➡️

MAI-Voice-2.1-Flash from @MicrosoftAI is live on OpenRouter. MAI-Voice-2.1-Flash is optimized for speed and scale: • Low-latency generation. Can generate 45s of audio within 150ms. • Natural, expressive speech with human-like intonation. • Same voice can speak in 23 languages while maintaining native accent. • $15 per 1M characters Built for voice agents, assistants, call centers, and other real-time experiences.

Grok Voice Think Fast 2.0 is STILL on top It ranks #1# on Artificial Analysis’ Speech Agent Arena with a 94.6% Task Success Rate Ahead of Gemini 3.8 Live, GPT-Realtime-2.1 High, GPT-Live-1 Astra and every other voice model tested And this benchmark measures real people talking to hidden voice agents across practical tasks... Then the benchmark checks whether the AI understood the request, used the right tools, and actually completed the job Not just sounding human but actually getting things done and Grok Voice is #1#

Conversational voice agents need to know when a different person is speaking, but not every background voice should count. We combined speaker embeddings with a post-trained audio-language model to determine when a speaker change matters.

Grok Voice Think Fast 2.0 is now #1# on the Artificial Analysis Speech-to-Speech Index. This index measures whether voice agents can reason over the speech it hears, resolve real customer issues, and correctly complete tasks using agent tools.