Instructions to use jubranAi/jubran-models with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use jubranAi/jubran-models 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 jubranAi/jubran-models:IQ2_S
# Run inference directly in the terminal:
llama cli -hf jubranAi/jubran-models:IQ2_S

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf jubranAi/jubran-models:IQ2_S
# Run inference directly in the terminal:
llama cli -hf jubranAi/jubran-models:IQ2_S

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf jubranAi/jubran-models:IQ2_S
# Run inference directly in the terminal:
./build/bin/llama cli -hf jubranAi/jubran-models:IQ2_S

Use Docker

docker model run hf.co/jubranAi/jubran-models:IQ2_S

Llama new
LM Studio
Jan
Ollama
How to use jubranAi/jubran-models with Ollama:
```
ollama run hf.co/jubranAi/jubran-models:IQ2_S
```
Unsloth Desktop

How to use jubranAi/jubran-models with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf jubranAi/jubran-models:IQ2_S

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": "jubranAi/jubran-models:IQ2_S"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use jubranAi/jubran-models with Docker Model Runner:
```
docker model run hf.co/jubranAi/jubran-models:IQ2_S
```

Lemonade

How to use jubranAi/jubran-models with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull jubranAi/jubran-models:IQ2_S

Run and chat with the model

lemonade run user.jubran-models-IQ2_S

List all available models

lemonade list

Hermes Agent

How to use jubranAi/jubran-models with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf jubranAi/jubran-models:IQ2_S

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 jubranAi/jubran-models:IQ2_S

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use jubranAi/jubran-models with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf jubranAi/jubran-models:IQ2_S

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 "jubranAi/jubran-models:IQ2_S" \
  --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"

Jubran model add-ons

The optional models the Jubran desktop app downloads after install, so the installer stays small. The app checks each file's SHA-256 (listed in SHA256SUMS and carried inside the app) before it uses it.

File	What it is	Size
`Qwen3-8B-Q4_K_M.gguf`	Qwen3-8B, 4-bit (Q4_K_M): the quick answering model	5.03 GB
`Qwen3-30B-A3B-Thinking-2507-IQ2_S.gguf`	Qwen3-30B-A3B-Thinking-2507, 2-bit (IQ2_S): the "Think deeper" model, about 3B parameters active per token	8.74 GB
`bge-reranker-base.tar`	BAAI/bge-reranker-base snapshot (optional; not used by the current app)	1.1 GB

Provenance and licences

Qwen3-8B and Qwen3-30B-A3B-Thinking-2507 are by the Qwen team, released under Apache-2.0. The GGUF conversions were made by third parties (the Qwen GGUF release for the 8B; the community IQ2_S quantisation for the 30B): before publishing, confirm the exact source repositories and that their licences allow redistribution, and name them here.
BAAI/bge-reranker-base is by BAAI, MIT licence.
The 30B file is a 2-bit quantisation: smaller and faster, and measurably less accurate than the original (see docs/deep_thinker_16gb.md in the Jubran repository).

Measured on an Apple M1, 16 GB

Quick tier (8B): 32/40 on a hard-question set, about 20 s per answer. Deep tier (30B): 35/40, about a minute. One run each on author-written questions; treat the difference as indicative.

Downloads last month: 61

GGUF

Model size

31B params

Architecture

qwen3moe

Hardware compatibility

2-bit

4-bit

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for jubranAi/jubran-models

Base model

BAAI/bge-reranker-base

Quantized

(26)

this model