Instructions to use IlyaasK/countdown-qwen3.5-0.8b-grpo with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use IlyaasK/countdown-qwen3.5-0.8b-grpo 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 IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M

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 IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M

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 IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M

Use Docker

docker model run hf.co/IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M

LM Studio
Jan
Ollama
How to use IlyaasK/countdown-qwen3.5-0.8b-grpo with Ollama:
```
ollama run hf.co/IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M
```
Unsloth Desktop

How to use IlyaasK/countdown-qwen3.5-0.8b-grpo with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M

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": "IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use IlyaasK/countdown-qwen3.5-0.8b-grpo with Docker Model Runner:
```
docker model run hf.co/IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M
```

Lemonade

How to use IlyaasK/countdown-qwen3.5-0.8b-grpo with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M

Run and chat with the model

lemonade run user.countdown-qwen3.5-0.8b-grpo-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use IlyaasK/countdown-qwen3.5-0.8b-grpo with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M

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 IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use IlyaasK/countdown-qwen3.5-0.8b-grpo with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M

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 "IlyaasK/countdown-qwen3.5-0.8b-grpo:Q4_K_M" \
  --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"

Qwen3.5-0.8B-Base Countdown GRPO · browser GGUF

This is a text-only, Q4_K_M GGUF export of a Qwen3.5-0.8B-Base checkpoint after 100,000 GRPO training steps on Jiayi Pan's Countdown-Tasks-3to4 arithmetic puzzles. No supervised fine-tuning was used. The vision tower and MTP weights are excluded from this browser export. The GGUF was converted with llama.cpp --no-nextn and quantized to Q4_K_M.

The model was trained to produce a single <answer>arithmetic expression</answer> using each supplied number once. It is experimental and often wrong. On the held-out evaluation used for this project, the final checkpoint solved 56/252 four-number puzzles in one attempt and 0/260 three-number puzzles. The browser demo defaults to four numbers and verifies every generated expression using exact rational arithmetic. It does not replace a failed answer with a solver-generated one.

Source base model: Qwen/Qwen3.5-0.8B-Base. Dataset: Jiayi-Pan/Countdown-Tasks-3to4. Training and demo source. Live demo.

Single-file download: countdown-Q4_K_M.gguf (SHA-256 fd5bae67451d67f6513b7563aa388561097bf347a774764404b4698b0c474cfe). The browser demo loads the equivalent five-part countdown-Q4_K_M-00001-of-00005.gguf split, which wllama downloads in parallel and reassembles in memory.

Downloads last month: 82

GGUF

Model size

0.8B params

Architecture

qwen35

Hardware compatibility

4-bit

Video Preview

Reinforcement Learning

Model tree for IlyaasK/countdown-qwen3.5-0.8b-grpo

Base model

Qwen/Qwen3.5-0.8B-Base

Quantized

(51)

this model