Deepgram's Nova-3 has expanded its enterprise-grade voice AI capabilities to include Italian, Turkish, Norwegian, and Indonesian, enhancing its linguistic diversity and adaptability. This expansion caters to a broad spectrum of speech structures and phonetic patterns, demonstrating Nova-3's ability to handle the complexities of different grammatical systems and regional variations. The update marks a significant advancement from previous iterations, enhancing accuracy and reducing Word Error Rate (WER) across batch and streaming modes, with streaming models showing the most significant gains. The inclusion of these languages opens new markets, providing enterprises with reliable voice AI solutions essential for sectors like banking, customer service, and tech ecosystems in Europe and Asia. Nova-3's Keyterm Prompting feature allows for seamless customization to accommodate domain-specific terms, further improving transcription precision. As Deepgram continues to expand Nova-3's capabilities, it positions itself as a robust ASR foundation for multilingual products and services, supporting dynamic real-time environments with improved latency and reduced errors.