Qwen3 0.6B 4-bit Core AI example

This repository contains a macOS Core AI .aimodel export of Qwen/Qwen3-0.6B, prepared for Apple-silicon Golden Gate (macOS 27) examples in MeshLLM.

It is a converted resource bundle, not the original SafeTensors checkpoint. The artifact uses Apple's 4-bit weight-only export with per-block block size 32 and an 8,192-token context limit. It requires Apple's Core AI runtime and cannot be loaded by llama.cpp, Transformers, or non-Apple hosts.

Use with MeshLLM

Download the repository on an Apple-silicon Mac with Golden Gate and point the experimental Apple provider at the downloaded resource directory:

hf download meshllm/qwen3-0.6b-4bit-aimodel \
  --local-dir target/apple-runtime/models/qwen3-0.6b-4bit-aimodel

export MESH_APPLE_COREAI_MODEL_ROOT="$PWD/target/apple-runtime/models/qwen3-0.6b-4bit-aimodel"
export MESH_APPLE_COREAI_MODEL_ID="apple/coreai/qwen3-0.6b"
export MESH_APPLE_COREAI_MODEL_VERSION="qwen3-0.6b-4bit-coreai"
export MESH_APPLE_COREAI_CONTEXT_SIZE=8192
export MESH_APPLE_COREAI_LANGUAGES=en
just apple::run serve --port 11435

The preparation manifest records the exact source revision, exporter revision, toolchain versions, and SHA-256 inventories used to produce this bundle.

Attribution and license

The source model is Qwen3 0.6B by Qwen and is distributed under the Apache License 2.0. The source revision used for this export is c1899de289a04d12100db370d81485cdf75e47ca. See LICENSE and the upstream model card for the full terms.

This converted artifact is provided as an experimental interoperability example. Apple Core AI and macOS are trademarks of Apple Inc.; this repository is not an Apple product or endorsement.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for meshllm/qwen3-0.6b-4bit-aimodel

Finetuned
Qwen/Qwen3-0.6B
Finetuned
(1188)
this model