Alma, a Voice LLM That's 63% Faster and 84% Cheaper Than OpenAI (finance.yahoo.com)

Phonely has unveiled Alma, a voice-specific language model (LLM) that significantly outperforms OpenAI's GPT-4.1 in both speed and cost efficiency. Trained on over 10 million real phone conversations, Alma delivers responses in under 200 milliseconds, compared to GPT-4.1's roughly 500 milliseconds, and costs only 55 cents per million tokens, making it 84% cheaper. This model is tailored for the nuanced nature of voice interactions, addressing challenges like interruptions and transcription errors, thereby enhancing the quality of call conversations. Alma's design allows for continuous improvement as it learns from live interactions, enabling a self-improving system that optimizes performance on the fly. As businesses conduct approximately $2.7 billion in revenue daily via phone calls, the efficiency and reliability improvements offered by Alma could greatly enhance sales conversion rates. Phonely has already seen substantial success, with one customer reporting a 3% increase in sales conversions translating into an additional $300,000 in monthly revenue. This technology represents a crucial advancement in bridging the gap between text-based AI models and the real-world demands of voice communication.