Liquid AI

Liquid AI describes a fraud-prevention deployment for an unnamed global payment processor and estimates 230 million dollars in additional fraud losses prevented annually, with on-premises or tenant deployment options.

Liquid AI reports 57 LFMs shipped, 48.2M downloads, 3,500+ variants and deployments across phones, laptops, cars, e-commerce, financial services, biology and defense, including Mercedes-Benz and Qualcomm partnerships.

Liquid AI reports 57 LFMs shipped, 48.2M downloads, 3,500+ variants and deployments across phones, laptops, cars, e-commerce, financial services, biology and defense, including Mercedes-Benz and Qualcomm partnerships.

Liquid AI develops compact device-native foundation models and novel sequence architectures optimized for edge, multimodal and agentic workloads.

Liquid AI develops compact device-native foundation models and novel sequence architectures optimized for edge, multimodal and agentic workloads.

Liquid AI extended LFM2.5 pretraining from 10T to 28T tokens, scaled post-training with reinforcement learning, and released Base, Instruct, Japanese, Vision-Language and Audio-Language models optimized for reliable on-device agents.

Liquid AI has built a cross-silicon and enterprise partner ecosystem spanning Mercedes-Benz, Qualcomm, AMD, Intel, Together AI and Modal.

Liquid AI has built a cross-silicon and enterprise partner ecosystem spanning Mercedes-Benz, Qualcomm, AMD, Intel, Together AI and Modal.

Liquid AI raised $250M in Series A funding to scale its Liquid Foundation Models, compute infrastructure and enterprise deployment across edge, on-premises and other private AI environments, with strategic and institutional backing for long-term model and systems development.

Liquid AI research and engineering roles span ML data, GPU performance, distributed training, multimodal/audio, inference systems and GPU infrastructure, while applied roles cover post-training, vision, audio and recommendation systems; its models target CPU, GPU and NPU deployment from edge devices to cloud.

Foundation-model, inference, multimodal and edge-deployment skills transfer exceptionally broadly across automotive, devices, cloud and physical AI companies.

Foundation-model, inference, multimodal and edge-deployment skills transfer exceptionally broadly across automotive, devices, cloud and physical AI companies.