Q4NX is FastFlowLM's native packed-quantization format - in the case of GPT-OSS a rearranged Q4_1-Q4NX layout tuned for the NPU matrix engine's tile sizes and memory access patterns. It is not a GGUF file and it does not run on llama.cpp or Ollama; it is meant exclusively for the FastFlowLM engine on AMD Ryzen AI NPUs.

Requirements

FastFlowLM >= 0.9.45 (flm CLI)
AMD Ryzen AI processor with XDNA2 (NPU2) - Strix Point / Ryzen AI 300 series or later
Linux with the XRT NPU stack installed
~16 GB of unified system memory (Q4NX weights + activations + KV cache)

Files

File	Purpose
model.q4nx	Quantized Q4NX weights
config.json	FastFlowLM model configuration
tokenizer.json	Tokenizer
tokenizer_config.json	Special tokens and chat template
chat_template.jinja	Chat template (optional)
flm-add.py	Installer script - registers this model with FastFlowLM

Install and run

This repository works with flm-add, a small installer that copies the model into the FastFlowLM user directory and registers the tag minicpm4.6:0.8b. It never modifies the system FastFlowLM install.

pip install flm-add or uv tool install flm-add

bash

uv tool install flm-add
flm-add Atomic-Germ/OpenAI-NEO-CODE-20B-NPU2 --tag openai-coder-neo:20b --family gpt-oss
FLM_CONFIG_PATH="$HOME/.config/flm/model_list.json" FLM_XCLBIN_PATH="$HOME/.config/flm" flm run minicpm-4.6:0.8b

Kernels

FastFlowLM's NPU kernels (xclbins) are closed source and are not shipped in this repository. flm-add.py links the kernels of the official gpt-oss:8b model (GPT-OSS-20B-NPU2), because this model shares the same engine family (gpt-oss) and architecture.

Serve (OpenAI-compatible)

flm serve openai-coder-neo:20b --port 8080

bash

curl http://127.0.0.1:8080/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{"model":"openai-neo-code:20b","messages":[{"role":"user","content":"Hello!"}],"max_tokens":256}'

Model

Registry tag: openai-coder-neo:20b
Engine family: gpt-oss
Kernel source: GPT-OSS-20B-NPU2
Context length: 131,072 tokens (from config)
model.q4nx size: 14.45 GB
Base model: openai/gpt-oss-20b
License: apache-2.0

Original model card

See the upstream model card for training details, benchmarks, and upstream usage. This repository only contains the Q4NX conversion for FastFlowLM.

Upstream card: Openai_gpt-oss-20b-CODER-NEO-CODE-DI-MATRIX-GGUF