Index-Translate-9B-q4f16_1-MLC

IndexTeam/Index-Translate-9B converted to the MLC-LLM format (q4f16_1) for in-browser inference with WebLLM / WebGPU.

• Weights: converted with mlc_llm convert_weight --quantization q4f16_1
• Config: mlc_llm gen_config --conv-template qwen2 --context-window-size 4096 (default system prompt removed, matching the official client)
• libs/Qwen3.5-9B-q4f16_1_cs1k-webgpu.wasm: WebGPU model library. Index-Translate-9B has exactly the same architecture and shapes as Qwen3.5-9B, so the official prebuilt library from mlc-ai/binary-mlc-llm-libs (v0_2_86) is reused unchanged.

WebLLM usage

import { CreateMLCEngine } from "@mlc-ai/web-llm";

const modelId = "Index-Translate-9B-q4f16_1-MLC";
const engine = await CreateMLCEngine(modelId, {
  appConfig: {
    model_list: [
      {
        model: "https://huggingface.co/chaceday/Index-Translate-9B-q4f16_1-MLC",
        model_id: modelId,
        model_lib:
          "https://huggingface.co/chaceday/Index-Translate-9B-q4f16_1-MLC/resolve/main/libs/Qwen3.5-9B-q4f16_1_cs1k-webgpu.wasm",
        overrides: { context_window_size: 4096 },
      },
    ],
  },
});

License

Apache-2.0, same as the original model. All credit goes to the IndexTeam authors (built on Qwen3.5).