Neuralwatt: Energy-based pricing for AI inference. Efficient prompts cost less (portal.neuralwatt.com)

Neuralwatt has introduced a groundbreaking AI inference API that shifts the pricing model from token-based to energy-based, allowing users to pay per kilowatt-hour and gain clear insights into their AI workloads' energy consumption. This innovative approach addresses the common challenge of hidden costs in traditional models, ensuring cost predictability and transparency. With features like real-time energy metrics per request and a comprehensive dashboard for usage trends and model comparisons, Neuralwatt empowers users to optimize their AI performance and efficiency. The platform leverages advanced technology, including vLLM architecture for state-of-the-art inference speed and multi-GPU tensor parallelism. It's designed for both hosted and on-premise use, offering significant improvements in energy efficiency—up to 40% more per computation. Neuralwatt aims to provide flexibility with an OpenAI-compatible API and a unique Power Optimization Engine that fine-tunes power consumption with minimal performance overhead. This development is significant for the AI/ML community as it not only promotes sustainable AI practices but also enables organizations to make informed decisions based on energy consumption, thereby leading to more efficient and cost-effective AI deployments.