Web AI Summit 2026 is an in-person event by Google focused on running AI and ML models entirely on the client side in JavaScript. Attendees will hear from front-end leaders about the cutting edge of client-side AI and edge computing, including local LLMs, computer vision, and real-time browser-based applications. The conference features lightning talks, demo…
Programmed at Web, in alphabetical order.
The companies behind the program. Tiers as disclosed.
Selected from 20 sessions.
Efficient on-device inference requires hardware-specific GPU kernels, but maintaining optimized implementations across devices is becoming increasingly complex. At Hugging Face, we…
One year after its debut as Google’s WebGPU runtime, LiteRT.js has evolved from a lightweight .tflite model runner into a complete, client-side AI platform. In this session, discov…
Bring powerful LLM capabilities to the web with LiteRT-LM.js. Keep user data local and API costs low by running models like Gemma 4 and Qwen directly in the browser. Go beyond simp…
Join us for an update on the web’s Built-in AI and AI-centric compute APIs. We’ll cover the Prompt and Task APIs, explore new multimodal inputs, and discuss Chrome’s shift toward s…
WebLLM brings local AI to browsers using MLC-LLM and TVM. As browser AI advances, systems challenges increasingly lie beyond model kernels. This talk explores runtime optimization …
In this talk, we introduce Llamas on the Web, a WebGPU backend for llama.cpp enabling efficient LLM inference in the browser. We’ll discuss challenges in memory management, quantiz…
One year after the v4 preview, Transformers.js has grown from 1.7 to 10 million monthly downloads. We cover what shipped: the new WebGPU runtime for browsers and server-side JS, mo…
For almost half a century, multi-camera analytics has been the gold standard of biomechanics, used in the most sophisticated labs around the world. Today, Factorial Biomechanics b…
I’ll show how I adapted image generation models for in-browser inference under strict constraints, from a 48 MB GAN implementation to a larger 400 MB diffusion pipeline. I’ll cover…
What does it mean when 19 companies show up, and 3 bet on three roles at once?
The numbers below are derived from the speakers, sponsors, and exhibitors on this page, cross-referenced into one ledger. They are the only thing here that 10times.com cannot tell you.