French startup ZML releases free inference server to end Nvidia lock-in

zml released llmd, a free inference server designed to optimize open-source large language models across a diverse range of hardware, including amd, google tpu, apple metal, and intel arc. founded by steeve morin and backed by yann lecun, the french startup targets the inefficiency of vendor lock-in, allowing companies to maximize chip performance and reduce soaring ai inference costs. although free initially, the proprietary software competes with established players like baseten and projects like vllm by offering broad hardware compatibility and co-design capabilities with silicon manufacturers.