wllama V3 and openjev bring local LLMs to the browser

Developer ngxson released wllama V3, llama.cpp's WebAssembly bindings with WebGPU and multimodal support, enabling fully in-browser LLM inference. Built on it, openjev and its SemIf experiment run local semantic if-decisions with no backend at all.

2026-09-19 ~ 2026-09-19 · 3 related posts

• openjev runs open LLMs locally in the browser via wllama (llama.cpp WebGPU) — ngxson · 2026-09-19
• SemIf: semantic 'if' decisions with local open models, entirely in your browser — ngxson · 2026-09-19
• wllama V3 ships WebGPU, multimodal and tool calling for in-browser llama.cpp inference — ngxson · 2026-09-19