We are thrilled to join forces with the Argmax team to bring frontier AI models to the edge. Argmax converges our vision of Responsible AI and Global Resilience to drive AI adoption in critical industries of our economy and society through high trust, high integrity open source infrastructure.
Despite recent advancements in AI, companies deploying frontier models in production are still largely constrained to server-side inference. Large models are too large to run on commodity hardware, while today’s compression techniques currently yield models below production-grade quality. This presents several hurdles for developers and enterprises looking to deploy foundation models in their products, including non-zero marginal cost for inference, privacy and security, latency and connectivity.
And yet, similar to how web developers capitalized on Electron to harness client hardware and build more sophisticated applications, we are witnessing growing enthusiasm within the open source AI developer community to bring the latest AI models to production on-device, promising higher performance and lower cost, and unlocking a new wave of innovative applications companies.
Empowering AI developers and enterprises to deploy and run large production-grade models on commodity hardware is the mission of Argmax, an open source developer platform pioneering the next generation of compression techniques and on-device inference software.
Few engineers are brave enough to delve into the intricacies of low-level hardware primitives. Even fewer perform the optimizations that enable a seamless experience on our everyday devices. And only a rare breed can innovate to push the boundaries tenfold. We are thrilled to partner with a remarkable trio of individuals that embody just that. Atila Orhon and his founding team members Brian Keene and Zach Nagengast not only have the talent, expertise and mindset that set them apart to take on this challenge, they are also fiercely ambitious, curious and authentic – the type of people we dream of partnering with.
Argmax leverages the founding team’s many years of experience building industry-leading inference software on Apple platforms and their major open source projects. These include Apple’s Neural Engine Transformers, the fastest mobile implementation of Stable Diffusion and key architecture contributions to Apple’s Core ML private inference engine.
The WhisperKit release is the first step toward an integrated suite of cutting-edge on-device inference performance and model lifecycle products that will simplify AI deployment for developers globally. We congratulate the entire Argmax team on their Seed round and are thrilled to welcome them to the GC family.
