Rime launches Prism speech-to-speech model for enterprises after $35 million raise

The San Francisco voice AI firm says its fine-tunable model beat OpenAI and Google on customer service benchmarks Rime, a San Francisco enterprise voice AI company, has built Prism, a speech-to-speech model aimed at businesses where a wrong answer on a phone call actually costs something. The launch follows a fundraising haul that the company’s announcement on X puts at $35 million. Prism was unveiled on October 28, 2026. Rime describes it as the first fine-tunable speech-to-speech model built specifically for enterprises. Speech-to-speech, or S2S, means the model takes in audio and produces audio, without the traditional relay race of separate transcription, reasoning, and voice systems. The twist is a checkpoint in the middle. Before Prism speaks, it generates a text output that can be verified. That gives companies a written record of what the model intends to say before a customer hears it. Prism can also be fine-tuned on a client’s own data. Companies can deploy it in a private cloud or fully on-premises, which addresses the compliance and security demands that keep many regulated firms away from public AI services. The numbers Rime is pointing to Speed is the first selling po...