Liquid AI Integrations

The LEAP Edge AI Platform provides an all-encompassing on-device AI toolchain enabling developers to construct edge AI applications, covering aspects from model selection to direct inference on the device itself. This innovative platform includes a best-model search engine that efficiently identifies the ideal model tailored to specific tasks and hardware constraints, alongside a variety of pre-trained model bundles available for quick download. Furthermore, it offers fine-tuning capabilities, complete with GPU-optimized scripts, allowing for the customization of models such as LFM2 to meet specific application needs. With its support for vision-enabled features across multiple platforms including iOS, Android, and laptops, the platform also integrates function-calling capabilities that enable AI models to interact with external systems via structured outputs. For effortless deployment, LEAP provides an Edge SDK that allows developers to load and query models locally, simulating cloud API functions while working completely offline. Additionally, its model bundling service simplifies the process of packaging any compatible model or checkpoint into an optimized bundle for edge deployment. This extensive array of tools guarantees that developers are well-equipped to efficiently and effectively build and launch advanced AI applications, ensuring a streamlined development process that caters to modern technological demands.

Apollo is an innovative mobile app that enables AI interactions entirely on-device, independent of cloud services, which allows users to engage with advanced language and vision models in a secure and private way with minimal latency. This application boasts a diverse array of compact foundation models drawn from the company's LEAP platform, empowering users to draft messages, send emails, interact with a personal AI assistant, create digital characters, and leverage image-to-text capabilities, all while functioning offline and ensuring that no data leaves the device. With a strong emphasis on instant responsiveness and offline operation, Apollo ensures that all processing occurs locally, removing the necessity for API calls, external servers, or the recording of user information. Serving as both a personal AI exploration tool and a development platform for those working with LEAP models, Apollo allows users to thoroughly evaluate a model's efficiency on their individual mobile devices before considering broader deployment. Furthermore, the application's design promotes user control and privacy, creating a smooth experience devoid of external disruptions and safeguarding personal data at every level. By prioritizing these aspects, Apollo not only enhances user trust but also encourages a more engaging interaction with AI technology.

SF Compute operates as a marketplace that provides users with on-demand access to vast GPU clusters, allowing for the rental of high-performance computing resources by the hour without requiring long-term contracts or significant upfront costs. Users can choose between virtual machine nodes or Kubernetes clusters that feature InfiniBand for quick data transfers, enabling them to specify the number of GPUs, the duration of use, and the start time based on their individual needs. The platform allows for customizable "buy blocks" of computing power; for example, clients may opt for a package of 256 NVIDIA H100 GPUs for three days at a set hourly rate, or they can modify their resource allocation to fit their financial plans. Kubernetes clusters can be deployed in just half a second, while virtual machines typically take around five minutes to be ready for use. In addition, SF Compute provides significant storage capabilities, boasting over 1.5 TB of NVMe and more than 1 TB of RAM, and users benefit from zero costs associated with data transfers in or out, ensuring no extra fees for data movement. The architecture of SF Compute cleverly obscures the physical infrastructure, utilizing a real-time spot market alongside a dynamic scheduling system to enhance resource allocation efficiency. This innovative arrangement not only improves usability but also significantly optimizes efficiency for clients aiming to expand their computational capacities, making it an attractive solution for various computing needs. Consequently, SF Compute stands out in the market by offering flexibility and cost-effectiveness that traditional computing solutions often lack.

The LFM-40B achieves a groundbreaking balance between model size and output quality. With 12 billion active parameters, it offers performance comparable to that of much larger models. Additionally, its mixture of experts (MoE) architecture significantly boosts throughput efficiency, making it ideal for use on cost-effective hardware. This unique blend of capabilities ensures remarkable results while minimizing the need for substantial resources. The design strategy behind this model emphasizes accessibility, allowing a wider range of users to benefit from advanced AI technology.