 | Built a single chatbot HTML page using Gemma 4 26B A4B running locally sharded between my 7900 XT and 3060 Ti with 32K context window at 50-65 t/s. Connects to LM Studio's API with full streaming, Markdown rendering, model selector, 6 parameter sliders, message editing with history branching, regenerate, abort, and system prompt support. Claude helped fix two DOM bugs that Gemma couldn't. Everything else was Gemma 4. GitHub: https://github.com/Shoggoth43/Gemma-4-26B-A4B-Generations [link] [comments] |