voice agent

"OpenAI announced in April 2026 that the Realtime API has been upgraded to General Availability (GA), officially releasing the new model gpt-realtime. Compared to the traditional preview version, there are three major changes: SIP phone support, Remote MCP server support, and asynchronous function calling."

"GPT-Realtime leads on Pass@1 (0.600) and interruption avoidance (13.5%); Gemini Live 3.1 achieves the fastest Latency (4.25 s) but the lowest turn-take rate (78.0%)."

"Deepgram’s STT and TTS model lineup now runs natively on Together AI... so teams can pair Deepgram transcription and synthesis with any LLM in the Together catalog and run the full voice pipeline on one production platform."

"Mistral releases Voxtral TTS model that’s fast, multilingual and small enough to be practical for voice agents."

""Gemini 3.1 Flash Live" is here, making AI audio sound more natural and reliable."

"According to Bland, building voice AI agents is extremely difficult and requires considerable expertise."

"I want to hear from people who’ve actually closed deals — not just “post on LinkedIn and pray” advice."

"So I’m building him an AI receptionist."

"Together AI, the AI Native Cloud, launched a full suite of capabilities to deliver the industry’s first unified solution for building real-time voice agents on co-located STT, LLM, and TTS infrastructure."

"The smallest model is less than 25 MB, and around 14M parameters."

"We present LALM-as-a-Judge, the first controlled benchmark and systematic study of large audio-language models (LALMs) as safety judges for multi-turn spoken dialogues."

"TTS-1.5 is designed to be more expressive and more stable than prior generations..."

"By working with strict latency […], the tutorial offers a valuable insight into optimizing performance."