Instructions to use Baekpica/Qwen3.8-Flash-Next-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use Baekpica/Qwen3.8-Flash-Next-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16
# Run inference directly in the terminal:
llama cli -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16
# Run inference directly in the terminal:
llama cli -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16
# Run inference directly in the terminal:
./llama-cli -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

Use Docker

docker model run hf.co/Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

LM Studio
Jan

vLLM

How to use Baekpica/Qwen3.8-Flash-Next-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Baekpica/Qwen3.8-Flash-Next-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Baekpica/Qwen3.8-Flash-Next-GGUF",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

Ollama
How to use Baekpica/Qwen3.8-Flash-Next-GGUF with Ollama:
```
ollama run hf.co/Baekpica/Qwen3.8-Flash-Next-GGUF:BF16
```
Unsloth Desktop

How to use Baekpica/Qwen3.8-Flash-Next-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "Baekpica/Qwen3.8-Flash-Next-GGUF:BF16"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use Baekpica/Qwen3.8-Flash-Next-GGUF with Docker Model Runner:
```
docker model run hf.co/Baekpica/Qwen3.8-Flash-Next-GGUF:BF16
```

Lemonade

How to use Baekpica/Qwen3.8-Flash-Next-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

Run and chat with the model

lemonade run user.Qwen3.8-Flash-Next-GGUF-BF16

List all available models

lemonade list

Hermes Agent

How to use Baekpica/Qwen3.8-Flash-Next-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use Baekpica/Qwen3.8-Flash-Next-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf Baekpica/Qwen3.8-Flash-Next-GGUF:BF16

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "Baekpica/Qwen3.8-Flash-Next-GGUF:BF16" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Qwen3.8-Flash-Next GGUF

Support my work

I work on making large language models practical on hardware they were never really designed to fit on — through mixed quantization, inference optimization, custom kernels, and serving experiments.

While much of the development happens on local hardware, calibration, profiling, and large-scale validation often require expensive on-demand GPUs.

Contributions help pay for that compute, storage, and testing infrastructure, so I can keep experimenting and publishing the results openly.

GGUF conversions of Qwen/Qwen3.8-Flash-Next, pinned to source revision f5d08274bafd880402bd16f5e3e6c514136ec06c.

Compatibility status

These files use a custom qwen4exp GGUF architecture prepared for the dfm branch of Baekpica/ds4, with DGX Spark/GB10 as the deployment target. Runtime support is still under implementation and validation. Do not assume compatibility with upstream llama.cpp or other GGUF runtimes unless they explicitly support this architecture and tensor schema.

Available variants

Variant	Status	Notes
`BF16`	verified / available	Lossless reference conversion. Source payload bits are preserved after semantic tensor splits.
`Q8_0`	verified / available	Most weight matrices use Q8_0; numerically sensitive or unsupported tensors remain BF16/F32/I64.

A separately tuned mixed-quant release will be published only after calibration, H200 quality checks, and ds4 runtime validation.

BF16 verification

The BF16 release contains 1,756 GGUF tensors in 12 shards, totaling 360,011,029,056 bytes. It was checked against all 1,658 tensors from the pinned source revision:

all 359,999,963,128 source payload bytes compared exactly;
zero payload mismatches;
fused expert gate/up tensors were split semantically without numerical conversion;
routed expert down projections were split losslessly into the main 512-column region and the 128-column tail;
GGUF shard metadata, tensor offsets, shapes, types, and Qwen Community License metadata were validated;
per-shard SHA-256 checksums are included alongside the files.

The source snapshot itself was also checksum-verified with hf cache verify before conversion.

Q8_0 verification

The Q8_0 release contains 1,756 GGUF tensors in 7 shards, totaling 192,201,208,384 bytes. Its audited tensor distribution is 806 Q8_0, 363 BF16, 584 F32, and 3 I64 tensors. GGUF metadata, shard numbering, tensor names, shapes, offsets, declared payload ends, and types were checked against the same exhaustive pinned source map with zero structural errors.

Per-shard SHA-256 checksums are included in Q8_0/SHA256SUMS. All seven public Hugging Face LFS object hashes and remote byte sizes were also compared with the local artifacts after upload and matched exactly.

Architecture notes

This conversion retains the complete multimodal and speculative-decoding topology: 48 text layers, 36 gated-delta layers, 12 full-attention layers, 512 routed experts with top-10 routing, shared experts, four hyper-connection streams, a 51.2B-parameter PLE n-gram table, the vision tower, and the MTP layer.

License

Use of these converted weights is governed by the original Qwen Community License 1.0. The exact upstream license file is included in this repository. No Apache-2.0 license is claimed for these weights.

Reproducibility

Conversion, verification, calibration, mixed-quant recipe, and ds4 runtime materials are being prepared for publication with the validated mixed-quant handoff. Until runtime validation is complete, the BF16 and Q8_0 files should be treated as conversion artifacts rather than a ready-to-run general-purpose release.

Downloads last month: 1,167

GGUF

Model size

180B params

Architecture

qwen4exp

Hardware compatibility

8-bit

16-bit

Model tree for Baekpica/Qwen3.8-Flash-Next-GGUF

Base model

Qwen/Qwen3.8-Flash-Next

Quantized

(171)

this model