gpt-oss-20b-MNN

gpt-oss-20b-MNN

MNN HQQ 4-bit inference package converted from openai/gpt-oss-20b.

Original model

OpenAI's gpt-oss-20b is an open-weight reasoning MoE model with 21B total parameters and 3.6B active parameters. It targets local and specialized applications with configurable reasoning effort and agentic capabilities. It uses the Harmony response format; preserve the bundled template and distinguish the final answer from reasoning or control channels.

Summary adapted from the upstream model card. Original model capabilities and benchmark claims are separate from the MNN smoke checks below.

Download

python -m pip install -U huggingface_hub
python -c "from huggingface_hub import snapshot_download; snapshot_download('taobao-mnn/gpt-oss-20b-MNN', local_dir='./gpt-oss-20b-MNN')"
export MODEL_DIR="$(pwd)/gpt-oss-20b-MNN"

MNN version and build

MNN: 7337404b2815b5fb9ddac13f4ac6a1fb37fcbbb4

This revision includes the MoE export fixes used for this release. Linux CPU build:

git clone https://github.com/alibaba/MNN.git
git -C MNN checkout 7337404b2815b5fb9ddac13f4ac6a1fb37fcbbb4
(cd MNN && ./schema/generate.sh)
cmake -S MNN -B MNN/build -DCMAKE_BUILD_TYPE=Release \
  -DMNN_BUILD_LLM=ON -DMNN_LOW_MEMORY=ON -DMNN_BUILD_CONVERTER=ON \
  -DMNN_SUPPORT_TRANSFORMER_FUSE=ON
cmake --build MNN/build --parallel 4

Conversion

The quantization arguments match export_args.json. Download the original checkpoint and export it with the revision above:

python -m pip install -r MNN/transformers/llm/export/requirements.txt
python -c "from huggingface_hub import snapshot_download; snapshot_download('openai/gpt-oss-20b', local_dir='./original/gpt-oss-20b', allow_patterns=['*.json', '*.safetensors', '*.model', '*.tiktoken', '*.jinja', '*.txt', '*.py'], ignore_patterns=['original/*', '*.gguf', '*.pt', '*.pth'])"

The original checkpoint uses MXFP4. Enable loading-time dequantization in its local configuration before HQQ export:

python - <<'PY'
import json
from pathlib import Path
path = Path("./original/gpt-oss-20b/config.json")
config = json.loads(path.read_text())
config["quantization_config"]["dequantize"] = True
path.write_text(json.dumps(config, indent=2))
PY

python MNN/transformers/llm/export/llmexport.py \
  --path ./original/gpt-oss-20b --export mnn --hqq \
  --quant_bit 4 --quant_block 64 --scale_bit 16 --embed_bit 16 \
  --lm_quant_bit 4 --lm_quant_block 64 \
  --mnnconvert ./MNN/build/MNNConvert --dst_path ./reexported/gpt-oss-20b-MNN

After re-export, reuse this package's verified tokenizer and template. Preserve the newly generated weight offsets and other model configuration:

python - <<'PY'
import json, os, shutil
from pathlib import Path
published = Path(os.environ["MODEL_DIR"])
converted = Path("./reexported/gpt-oss-20b-MNN")
for filename in ("llm_config.json", "config.json"):
    path = converted / filename
    config = json.loads(path.read_text())
    reference = json.loads((published / filename).read_text())
    for key in ("jinja", "use_template"):
        config.pop(key, None)
        if key in reference:
            config[key] = reference[key]
    if filename == "config.json":
        config["tokenizer_file"] = "tokenizer.mtok"
    path.write_text(json.dumps(config, ensure_ascii=False, indent=2))
shutil.copyfile(published / "tokenizer.mtok", converted / "tokenizer.mtok")
PY

Run and test

printf 'What is 2 + 3? Reply with only the number.\n' > prompt.txt
./MNN/build/llm_demo "$MODEL_DIR/config.json" prompt.txt
./MNN/build/llm_bench -m "$MODEL_DIR/config.json"

The original migration passed 6 CPU High/Low smoke cases for this model. These are functional checks, not a broad quality or GPU benchmark.

License and attribution

Weights retain the upstream license and attribution. See the original model card, the license files in this repository and the MNN LLM documentation.