ONNX Runtime Web
The browser-targeted variant of ONNX Runtime, Microsoft's cross-platform inference engine. It supports direct loading of standard ONNX-format models and is used for deploying AI inference workloads on the web client side.
Core Facts
ONNX Runtime Web compiles the ONNX runtime to WebAssembly to enable cross-platform model inference; ONNX was jointly introduced by Microsoft and Facebook.
Transformers.js is a JavaScript library developed by Hugging Face that, based on WebAssembly and WebGPU technology, ports Transformer model inference to browser environments.
Timeline (last 90 days)
Transformers.js底层使用ONNX Runtime Web，模型需要转为ONNX格式并量化，4-bit量化的1B模型大小约1.2GB而非5GB。
WASM后端1.6 tokens/s的低性能部分是因为页面未启用跨源隔离，导致ONNX Runtime Web使用单线程执行。
TaskProof's optional AI component uses the FLAN-T5 Small model, running inside a browser Web Worker via Transformers.js and ONNX Runtime Web, and its role is limited to summarizing possible deliverables, asking one question for the buyer, and drafting one opening line.
With the maturation of technologies such as WebGPU and ONNX Runtime Web, running small language models in the browser is becoming a reality.
By compiling a machine learning inference engine (such as ONNX Runtime Web or Transformers.js) into a Wasm module and loading quantized model weights, image inference can be completed locally on the user's device, with photos never leaving the local device.
Talus is exported in ONNX format with weights stored in fp16 and converted back to fp32 on load, runs on WebGPU via ONNX Runtime Web, generates one map in approximately 3 seconds on an RTX 5060, and provides a CPU fallback.
ONNX Runtime Web is the browser version of Microsoft's cross-platform inference engine, capable of directly loading standard ONNX format models
WebAssembly and WebGPU are two primary technical paths for in-browser AI inference, and frameworks such as ONNX Runtime Web and TensorFlow.js already support both approaches.
Modern browsers' Web Speech API provides a native speech recognition interface, and technologies such as WebGPU and ONNX Runtime Web allow lightweight neural network models to run directly in the browser.
Browser-side automatic background removal mainly relies on WebAssembly and browser-based machine learning frameworks such as TensorFlow.js or ONNX Runtime Web to load pre-trained image segmentation models.
1 more timeline events
All Facts (12)
ONNX Runtime Web compiles the ONNX runtime to WebAssembly to enable cross-platform model inference; ONNX was jointly introduced by Microsoft and Facebook.
75%VerifiedTransformers.js is a JavaScript library developed by Hugging Face that, based on WebAssembly and WebGPU technology, ports Transformer model inference to browser environments.
70%UnverifiedBrowser-side automatic background removal mainly relies on WebAssembly and browser-based machine learning frameworks such as TensorFlow.js or ONNX Runtime Web to load pre-trained image segmentation models.
60%UnverifiedTransformers.js底层使用ONNX Runtime Web，模型需要转为ONNX格式并量化，4-bit量化的1B模型大小约1.2GB而非5GB。
50%UnverifiedWASM后端1.6 tokens/s的低性能部分是因为页面未启用跨源隔离，导致ONNX Runtime Web使用单线程执行。
50%UnverifiedTaskProof's optional AI component uses the FLAN-T5 Small model, running inside a browser Web Worker via Transformers.js and ONNX Runtime Web, and its role is limited to summarizing possible deliverables, asking one question for the buyer, and drafting one opening line.
50%UnverifiedBy compiling a machine learning inference engine (such as ONNX Runtime Web or Transformers.js) into a Wasm module and loading quantized model weights, image inference can be completed locally on the user's device, with photos never leaving the local device.
50%UnverifiedTalus is exported in ONNX format with weights stored in fp16 and converted back to fp32 on load, runs on WebGPU via ONNX Runtime Web, generates one map in approximately 3 seconds on an RTX 5060, and provides a CPU fallback.
50%UnverifiedONNX Runtime Web is the browser version of Microsoft's cross-platform inference engine, capable of directly loading standard ONNX format models
50%UnverifiedWebAssembly and WebGPU are two primary technical paths for in-browser AI inference, and frameworks such as ONNX Runtime Web and TensorFlow.js already support both approaches.
50%UnverifiedModern browsers' Web Speech API provides a native speech recognition interface, and technologies such as WebGPU and ONNX Runtime Web allow lightweight neural network models to run directly in the browser.
50%UnverifiedWith the maturation of technologies such as WebGPU and ONNX Runtime Web, running small language models in the browser is becoming a reality.
50%