DictaLM-3.0-1.7B-Instruct — q4f16_1 (MLC / WebLLM)

Conversion of dicta-il/DictaLM-3.0-1.7B-Instruct (by Dicta, Apache-2.0) to the MLC q4f16_1 format, so it runs in the browser with WebLLM on WebGPU.

DictaLM 3.0 1.7B uses the Qwen3-1.7B architecture, so this conversion reuses WebLLM's prebuilt Qwen3-1.7B-q4f16_1_cs1k-webgpu.wasm model library. Weight layout and mlc-chat-config.json mirror mlc-ai/Qwen3-1.7B-q4f16_1-MLC; tokenizer files come from the original model.

const appConfig = {
  model_list: [{
    model: "https://huggingface.co/sivelsguy/DictaLM-3.0-1.7B-Instruct-q4f16_1-MLC",
    model_id: "DictaLM-3.0-1.7B-Instruct-q4f16_1-MLC",
    model_lib: "https://raw.githubusercontent.com/mlc-ai/binary-mlc-llm-libs/main/web-llm-models/v0_2_86/base/Qwen3-1.7B-q4f16_1_cs1k-webgpu.wasm",
    overrides: { context_window_size: 4096 },
  }],
};

Used by the open-source Hebrew learning app Parle. All credit for the model goes to Dicta.