Hot French startup ZML releases free product to speed inference across lots of AI chips

Paris-based startup ZML, supported by Yann LeCun, has introduced ZML/LLMD, an inference server software designed to run large language models efficiently across a diverse range of hardware, including chips from Nvidia, AMD, Google, Apple, and Intel. By enabling peak performance on various processors, the company aims to combat vendor lock-in and address rising AI costs. Although ZML faces competition from established players like vLLM and SGLang, its founder, Steeve Morin, emphasizes their unique approach to silicon co-design. Currently offered as a free product, the tool intends to foster enterprise adoption while helping the company gather usage data before moving toward a future revenue model.

(Source：TechCrunch)