Local runtime · Embedded library

WebLLM

In-browser LLM inference engine built on MLCEngine that runs entirely client-side through WebGPU, exposing an OpenAI-compatible JavaScript and TypeScript API with streaming, JSON-mode structured generation, and Web Worker, Service Worker, and browser-extension integration.

Maintainer

MLC AI community

Accelerators

WebGPU

Model formats

MLC compiled

Source model

Open source

Licenses

Apache License 2.0

Boundary

Represents the mlc-ai/web-llm package: its TypeScript API, prebuilt-model registry, and the browser-side MLCEngine that runs entirely inside a web page through WebGPU and WebAssembly, with no local server process. It is published, versioned, and licensed separately from mlc-ai/mlc-llm, the sibling record covering the Python compiler, CLI, and REST server; both consume the same compiled MLC model format, but a genuinely new model architecture still depends on mlc-llm's Python compilation pipeline to produce the WASM model library this package loads. WebLLM Chat (chat.webllm.ai) and the community web-llm-chat project are demonstration and third-party chat applications built on this engine, not part of the reviewed package, so it stays on the runtime side of the ADR 015 test.

Open in the directory → Open official documentation ↗ Repository ↗

Editorial ratings appear in the directory beside the profile they belong to and are never compared across collections.