Instructions to use basecompute/Qwen3.8-27B-BaseQuant-V1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use basecompute/Qwen3.8-27B-BaseQuant-V1 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 basecompute/Qwen3.8-27B-BaseQuant-V1
# Run inference directly in the terminal:
llama cli -hf basecompute/Qwen3.8-27B-BaseQuant-V1

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf basecompute/Qwen3.8-27B-BaseQuant-V1
# Run inference directly in the terminal:
llama cli -hf basecompute/Qwen3.8-27B-BaseQuant-V1

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 basecompute/Qwen3.8-27B-BaseQuant-V1
# Run inference directly in the terminal:
./llama-cli -hf basecompute/Qwen3.8-27B-BaseQuant-V1

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 basecompute/Qwen3.8-27B-BaseQuant-V1
# Run inference directly in the terminal:
./build/bin/llama-cli -hf basecompute/Qwen3.8-27B-BaseQuant-V1

Use Docker

docker model run hf.co/basecompute/Qwen3.8-27B-BaseQuant-V1

Llama new
LM Studio
Jan

vLLM

How to use basecompute/Qwen3.8-27B-BaseQuant-V1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "basecompute/Qwen3.8-27B-BaseQuant-V1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "basecompute/Qwen3.8-27B-BaseQuant-V1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/basecompute/Qwen3.8-27B-BaseQuant-V1

Ollama
How to use basecompute/Qwen3.8-27B-BaseQuant-V1 with Ollama:
```
ollama run hf.co/basecompute/Qwen3.8-27B-BaseQuant-V1
```
Unsloth Desktop

How to use basecompute/Qwen3.8-27B-BaseQuant-V1 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf basecompute/Qwen3.8-27B-BaseQuant-V1

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": "basecompute/Qwen3.8-27B-BaseQuant-V1"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use basecompute/Qwen3.8-27B-BaseQuant-V1 with Docker Model Runner:
```
docker model run hf.co/basecompute/Qwen3.8-27B-BaseQuant-V1
```

Lemonade

How to use basecompute/Qwen3.8-27B-BaseQuant-V1 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull basecompute/Qwen3.8-27B-BaseQuant-V1

Run and chat with the model

lemonade run user.Qwen3.8-27B-BaseQuant-V1-{{QUANT_TAG}}

List all available models

lemonade list

Hermes Agent

How to use basecompute/Qwen3.8-27B-BaseQuant-V1 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf basecompute/Qwen3.8-27B-BaseQuant-V1

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 basecompute/Qwen3.8-27B-BaseQuant-V1

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use basecompute/Qwen3.8-27B-BaseQuant-V1 with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf basecompute/Qwen3.8-27B-BaseQuant-V1

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 "basecompute/Qwen3.8-27B-BaseQuant-V1" \
  --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"

Add XS tier (8.36 GB, size-matched to unsloth UD-IQ2_S)

by prabodbc - opened 4 days ago

base: refs/heads/main

←

from: refs/pr/2

Discussion Files changed

+29

-19

prabodbc

Base Compute org 4 days ago

Adds an XS tier, the same size as unsloth UD-IQ2_S (8,371,970,048 B):

Qwen3.8-27B-BaseQuant-V1-XS.gguf: 8,363,981,344 B, sha256 a4b3b65c338e1304c55f309c53fc8266509fa480c3e9886aa298c111b32cb270
Qwen3.8-27B-BaseQuant-V1-XS.base: 8,362,584,064 B, sha256 49e69ca794e84807a53e9933b70bfbf314b3808c17a516757f4ba94cba29da4c
README.md: an XS row and XS columns in the result tables

Both files are text-only, with no MTP layer, like UD-IQ2_S.

Results (KL to BF16, unsloth / XS):

Held-out model continuations: 0.0914 / 0.0776, -15%, 95% CI of the difference [-0.0165, -0.0111]. 256 chunks.
Held-out written text: 0.8349 / 0.7308, -12%. 256 chunks.
Public prompts, all six sources: 0.1441 / 0.1242, -14% [-18%, -10%]. Top-1 88.5% / 89.1%.

The held-out rows come from the XS campaign's validation split. The tool-call rows were not measured for XS. The plots were not changed and still show only Small, Medium and Large.

baseRT bundle: it uses the GGUF's own blocks for 354 of 402 tensors, with the V-head rows of qkv and z moved back into HF order. The 48 linear-attention out_proj tensors were re-quantized in HF column order. On an M5 Max with baseRT 0.4.1, over 40 SWE-agent and OpenHands prompts (up to 6.7k tokens), KL is 0.1089 against llama.cpp XS's 0.1100.

Add XS tier (8.36 GB, size-matched to unsloth UD-IQ2_S)494ebb18

prabodbc changed pull request status to merged 3 days ago

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment