Back to Tool
W

webyIQ

Tool Nov 3, 2026

A community Rust and WebGPU inference engine specialized for small models, with browser chat demos and a Chrome or Edge DOM-agent extension.

Visit primary link 2 sources attached
Author-published webyIQ chat arena screenshot showing two engines answering a coffee-shop naming prompt, with displayed generation rates of 506.7 and 27.0 tokens per second.
Author-published chat-arena session. Its displayed 506.7 tok/s result is a separate demonstration, not the headline 620 tok/s run or an AITrending reproduction. Source: webyIQ / rtx527896828
agent-ecosystembrowser-automationself-hosted

What happened lately

webyIQ timeline

webyIQ specializes browser inference for Qwen3.5-2B and MiniCPM5-2B rather than using a general graph runtime. The author supplies a hosted chat demo, conversion tools, a local DOM-agent demo and an unpacked Chrome or Edge extension.

The November 2 repository commits document hosted weights and comparison pages; a November 3 English community post introduces the project. Performance figures are author measurements on specified hardware. The prototype was tested on Windows with Chrome and Edge 156; Safari and mobile remain untested. Its extension requests access to all sites to read and operate pages.