Voice AI startup Phonic announced a USD 27.5m Series A led by First Round Capital on October 29, 2026, bringing its total funding to USD 31.5m, with Lux Capital, Felicis, Definition, and Essence VC also participating. The same day, the San Francisco-based company launched Phonic-1, a speech-to-speech model for voice agents. Phonic has focused exclusively on speech-to-speech AI since its founding more than two years ago. Unlike cascaded voice AI systems such as those from Bland, ElevenLabs, and Deepgram, which chain together transcription, text-based response generation, and speech synthesis, Phonic-1 processes and generates speech directly within a single model. Phonic sits in the speech model layer of the voice AI ecosystem, supplying technology that enables voice agents to process and generate speech, and argues its approach preserves conversational cues such as tone, pauses, and the pronunciation of names. Competitors include OpenAI’s GPT-Live-1, Google’s Gemini 3.8 Live, and Amazon’s Nova Sonic. According to Phonic, its model was trained on more than 50 languages and can switch between languages during a conversation while maintaining the same speaker identity, and can begin using external tools while the caller is still speaking. Phonic said Phonic-1 achieved a score of 83.2 on a benchmark from Sierra, higher than OpenAI’s GPT-Live-1 and