# Voice agents

ElevenLabs and NVIDIA are expanding their partnership to bring NVIDIA's Nemotron open models into ElevenAgents, ElevenLabs' conversational agent platform, and to add 15 MW of Vera Rubin capacity on top of existing Blackwell Ultra systems. On the ElevenAgents benchmark for realistic customer-support tasks, Nemotron 3.5 Super VL reached leading end-to-end task-completion accuracy at 70% lower cost than other leading open models. The two companies are also applying voice agents to Poland's national e-Health central appointment system, with an on-premises deployment that keeps citizen data and audio within government infrastructure.

ElevenLabs and NVIDIA are expanding their partnership to bring NVIDIA's Nemotron open models into ElevenAgents, ElevenLabs' conversational agent platform, and to add 15 MW of Vera Rubin capacity on top of existing Blackwell Ultra systems. On the ElevenAgents benchmark for realistic customer-support tasks, Nemotron 3.5 Super VL reached leading end-to-end task-completion accuracy at 70% lower cost than other leading open models. The two companies are also applying voice agents to Poland's national e-Health central appointment system, with an on-premises deployment that keeps citizen data and audio within government infrastructure.

Vercel and ElevenLabs are partnering to bring 14 ElevenLabs audio models to AI Gateway, spanning speech to text, streaming text to speech, realtime voice agents, music, sound effects and multi-voice dialogue. Previously these capabilities had to be wired up separately; now an app or agent can transcribe a user's speech, reason over it with a language model of its choice and answer in an ElevenLabs voice through a single gateway API key, with every request landing in the same usage, cost and trace views as other AI Gateway traffic. All ElevenLabs models are 50% off on AI Gateway for the first 14 days, through November 9th.

Vercel and ElevenLabs are partnering to bring 14 ElevenLabs audio models to AI Gateway, spanning speech to text, streaming text to speech, realtime voice agents, music, sound effects and multi-voice dialogue. Previously these capabilities had to be wired up separately; now an app or agent can transcribe a user's speech, reason over it with a language model of its choice and answer in an ElevenLabs voice through a single gateway API key, with every request landing in the same usage, cost and trace views as other AI Gateway traffic. All ElevenLabs models are 50% off on AI Gateway for the first 14 days, through November 9th.

Mistral released Voxtral 3, which puts speech recognition, reasoning, tool calling and speech synthesis into a single real-time streaming pipeline, alongside a voice agent SDK. Previously, making an agent listen accurately, reason, speak naturally and act across systems in multiple steps remained difficult. Its Small transcription model beats Whisper, Scribe v2 and Gemini 3.5 Transcribe on multilingual transcription and diarization in English, French and Spanish, and realtime transcription runs below 500 ms latency.

Mistral released Voxtral 3, which puts speech recognition, reasoning, tool calling and speech synthesis into a single real-time streaming pipeline, alongside a voice agent SDK. Previously, making an agent listen accurately, reason, speak naturally and act across systems in multiple steps remained difficult. Its Small transcription model beats Whisper, Scribe v2 and Gemini 3.5 Transcribe on multilingual transcription and diarization in English, French and Spanish, and realtime transcription runs below 500 ms latency.

Salesforce AI Research has built Launa-Omni, a voice model for real-time enterprise conversation that keeps speech and language in one audio-adapted text model and uses a separate in-house text-to-speech model to speak the reply. Unlike pipelines that chain speech recognition, a language model and speech generation, speech is not transcribed before reasoning, so pace, hesitation and emphasis survive the first handoff. The model is significantly smaller than several larger voice models and reports competitive results on spoken instruction following, multi-turn dialogue and customer-service tasks; a simplified version already powers Slack's voice dictation.

Salesforce AI Research has built Launa-Omni, a voice model for real-time enterprise conversation that keeps speech and language in one audio-adapted text model and uses a separate in-house text-to-speech model to speak the reply. Unlike pipelines that chain speech recognition, a language model and speech generation, speech is not transcribed before reasoning, so pace, hesitation and emphasis survive the first handoff. The model is significantly smaller than several larger voice models and reports competitive results on spoken instruction following, multi-turn dialogue and customer-service tasks; a simplified version already powers Slack's voice dictation.

PhoneLLM Mini, a model from the Pipecat team at Daily, is now available on Cloudflare Workers AI for voice agents and other realtime AI applications. In a voice agent the language model sits between speech-to-text and text-to-speech, so every second it takes adds to the caller's wait; developers usually choose between a fast model without reasoning mode, which can miss instructions or tool calls, and a slower model in reasoning mode, which handles both well. Pipecat post-trained NVIDIA Nemotron 3 Nano 30B on phone and customer support scenarios so PhoneLLM follows instructions and calls tools better without reasoning mode.

PhoneLLM Mini, a model from the Pipecat team at Daily, is now available on Cloudflare Workers AI for voice agents and other realtime AI applications. In a voice agent the language model sits between speech-to-text and text-to-speech, so every second it takes adds to the caller's wait; developers usually choose between a fast model without reasoning mode, which can miss instructions or tool calls, and a slower model in reasoning mode, which handles both well. Pipecat post-trained NVIDIA Nemotron 3 Nano 30B on phone and customer support scenarios so PhoneLLM follows instructions and calls tools better without reasoning mode.

Sarvam released Bulbul V4 Flash, a text-to-speech model for Indian languages that reaches under 250 ms time-to-first-audio in production and costs ₹20 per 10,000 characters. That is a third less than V3's ₹30, while the company reports richer prosody, stronger contextual delivery and higher speech quality. The model launches with 83 voices across 11 languages and 224 voice-and-style combinations, plus a real-time contextual text normalization layer for items such as PNR, PAN and Aadhaar whose spoken form depends on context.

Sarvam released Bulbul V4 Flash, a text-to-speech model for Indian languages that reaches under 250 ms time-to-first-audio in production and costs ₹20 per 10,000 characters. That is a third less than V3's ₹30, while the company reports richer prosody, stronger contextual delivery and higher speech quality. The model launches with 83 voices across 11 languages and 224 voice-and-style combinations, plus a real-time contextual text normalization layer for items such as PNR, PAN and Aadhaar whose spoken form depends on context.

Gradium TTS pushes time-to-first-audio below typical human speech response, reaching a median TTFA of 214 ms on the Coval benchmark as of September 2026 with an interquartile range tight enough for developers to set reliable latency budgets for the rest of the pipeline. It uses Kyutai's Delayed Streams Modeling, which treats text and audio as two parallel streams on a shared clock so audio can start before the upstream model finishes the sentence. On Baseten it ships as a dedicated deployment whose weights are pre-mirrored and cached, cutting cold starts by 2–3x.

Gradium TTS pushes time-to-first-audio below typical human speech response, reaching a median TTFA of 214 ms on the Coval benchmark as of September 2026 with an interquartile range tight enough for developers to set reliable latency budgets for the rest of the pipeline. It uses Kyutai's Delayed Streams Modeling, which treats text and audio as two parallel streams on a shared clock so audio can start before the upstream model finishes the sentence. On Baseten it ships as a dedicated deployment whose weights are pre-mirrored and cached, cutting cold starts by 2–3x.