ONNX + Transformers.js: Run Models Like Liquid LFM in the Browser
Melvin Vivas · X video post · 2026-10-22 · 1:06 · 327 views · Open on X
Topics: LLMOps, Deployment & Monitoring, LLM Fundamentals · Level: beginner
Summary
This is a short interview clip about ONNX (Open Neural Network Exchange), an open standard that lets you train a model in one framework and run it in another. It explains that ONNX defines the operators a model needs, while hardware vendors such as AMD, Intel and Qualcomm build their own runtimes to make it fast. It also shows how Transformers.js can load an ONNX-exported model, like a Liquid Foundation Model (LFM), inside a single-page web app so it runs in the browser with no compiler or separate runtime.
Key points
- ONNX stands for Open Neural Network Exchange. It is an open standard for moving models between frameworks: you train in one and run inference in another.
- ONNX defines the set of operators a neural network needs to run inference. That makes a model portable.
- Speed depends on the hardware vendor. AMD, Intel and Qualcomm each provide their own runtime providers tuned for their chips.
- Transformers.js can load an ONNX-exported model directly in the browser.
- You can build a single-page app that loads the ONNX model and runs it right away, with no compiler or separate runtime to install.
- The example given is running a Liquid Foundation Model (LFM) in the browser through ONNX and Transformers.js.
Resources mentioned
- ONNX · tool · onnx.ai · free
An open standard format for neural networks that lets models trained in one framework run in other runtimes and on different hardware.
Also in: Building a Local Model Server with ONNX Support (Melvin Vivas on X · notes) - Transformers.js · tool · huggingface.co · free
Hugging Face's JavaScript library for running transformer models directly in the browser or in Node.js.
Also in: Real-Time Speech Transcription in the Browser with Voxtral and WebGPU (Melvin Vivas on X · notes) - Liquid AI (@liquidai) on X · website · x.com · free
Liquid AI's family of efficient foundation models, used here as the example of a model you can run in the browser with ONNX.
Also in: Trying Liquid AI Decision Models (d1, d1 ultra) via API (Melvin Vivas on X · notes), Liquid RIPL: Low-Latency Inference Engine for LFMs (Melvin Vivas on X · notes), Try Liquid AI's open d1-3B decision model in Colab (Melvin Vivas on X · notes), Liquid AI releases d1-ultra decision model (Melvin Vivas on X · notes) and 30 more
Try this
- Try Transformers.js: build a simple single-page app that loads an ONNX-exported model and runs it in the browser.
- Look up the ONNX runtime providers from your hardware vendor (AMD, Intel or Qualcomm) to get better performance.
- A single-page web app that runs a Liquid LFM (or another ONNX-exported model) entirely in the browser using Transformers.js.
More in LLMOps, Deployment & Monitoring
- Claude models (Opus 5.5) generally available on Google Cloud
- DeepInfra raising GPU prices
- Liquid RIPL: Low-Latency Inference Engine for LFMs
- Railway Platform Incident: Check Provider Status Pages
- Run Pi with local models on an RTX Spark laptop via llama.cpp
- Local AI vs Cloud AI: When Running Models Locally Makes Sense